TL;DR

• A TD Cowen note put a price on the carriers’ fear: SpaceX could buy T-Mobile to force its way into wireless. The financial-press version prices the spectrum (about $20 billion of EchoStar licenses around 2 GHz) and the rockets, then stops.

• A standard handset reaches a Starlink satellite because the satellite speaks standard cellular. SpaceX runs conventional Long Term Evolution (”LTE”) base stations in orbit, technology the 3rd Generation Partnership Project (”3GPP”) standardized and that carries declared Standard Essential Patents (”SEPs”), with its own proprietary satellite adaptations on top. The reach rides licensed technology SpaceX did not invent.

• Today’s direct-to-cell is Non-Terrestrial Network (”NTN”) bolted onto a terrestrial-first 5G design. 6G integrates the satellite layer from the start. Boston Consulting Group and Qualcomm describe 6G as unifying terrestrial and non-terrestrial layers.

• The 6 GHz band (radio spectrum, distinct from the 6G cellular generation) splits three ways. The United States made all 1200 megahertz unlicensed for Wi-Fi, China made all of it licensed mobile, and Europe, the United Kingdom and India split it. The 2027 World Radiocommunication Conference (”WRC-27”) decides the upper band.

• AI policy now treats open-source and royalty-free as the default path, and a narrower version has surfaced in standards policy: early proposals to open the specification process to royalty-free and open-source contributions. That lets a rival with a large engineering workforce and a state-coordinated standards strategy build on US-funded research without paying for it. Chinese open-weight models already lead the open-source field.

• Upgrade velocity is the edge the takeover-math misses. Software-defined satellites take new 3GPP releases by update, and the five-year orbit refreshes the hardware, so Starlink can roll the network to 6G faster than carriers can swap tower radios. A network that keeps implementing more of the standard carries a growing licensing obligation, not a one-time one.

• The National Telecommunications and Information Administration (”NTIA”) runs Mission LA 2028 to demonstrate early 6G at the Summer Olympic Games. The Presidential Memorandum “Winning the 6G Race” names the priority. A 6G strategy that celebrates well-funded implementers while casting contributors as the antitrust problem works against its own goal.

What the market saw

The Starlink mobile story comes in two versions. Wall Street prices a takeover. On 26 June 2026, TheStreet relayed a TD Cowen note from analyst Gregory Williams: if rivals keep blocking SpaceX from leasing capacity through a Mobile Virtual Network Operator (”MVNO”) deal, SpaceX could buy T-Mobile outright, a company carrying a market value near $250 billion. Oppenheimer told clients SpaceX could disrupt the $1.6 trillion US communications industry. The feeds carry the other: infographics for “Starlink Mobile, the first hybrid terrestrial-satellite network,” promising one device, one network, everywhere. (link)

The Federal Communications Commission (”FCC”) approved SpaceX’s purchase of EchoStar spectrum in 2026, roughly 65 megahertz (“MHz”) around 2 gigahertz (“GHz”)(the AWS-3, AWS-4 and H-Block bands) for about $19.6 billion, with a technology-neutral waiver to use it across terrestrial, satellite and hybrid networks. SpaceX counts more than 10 million Starlink subscribers. Verizon, AT&T and T-Mobile pooled spectrum in a satellite-to-phone venture in May 2026, read across the trade press as an effort to slow Starlink.

Both versions treat the same two assets as the barrier to entry: the spectrum and the constellation. Neither asks who licenses the technology that makes a hybrid network work. The hybrid is not new, and it is not SpaceX’s. It is what 3GPP standardized as Non-Terrestrial Network. The question both skip is how an unmodified iPhone reaches a satellite in the first place.

The answer is a standard, and the standard is licensed

SpaceX describes its newest satellites as “cell towers in space” that link ordinary smartphones with no special hardware. The phone connects because the satellite speaks standard LTE, the same 3GPP air interface the phone runs with a ground tower, and that interface carries thousands of declared SEPs. Starlink reached the market by putting conventional LTE base stations on its satellites with proprietary adaptations for the satellite link, ahead of and outside the 3GPP Release 17 satellite profiles (the New Radio and Internet of Things NTN profiles, frozen in 2022) that the broader industry and 6G are converging on. The licensed layer is the air interface to the handset. The satellite-side adaptation is SpaceX’s own. An unmodified phone reaches the satellite only because Starlink speaks a standard the phone already implements.

The routing runs in two standardized modes. In the transparent mode the satellite relays the signal to a ground gateway, and the base station sits on the ground. In the regenerative mode the base station runs onboard the satellite, which decodes traffic in orbit and can pass it between satellites over laser links before it reaches a ground anchor. A 2025 crowdsourced-measurement study found Starlink uses the regenerative path. Its Direct to Cell satellites carry an onboard eNodeB, a base station in orbit, and the Gen2 fleet uses inter-satellite laser links so traffic need not always route through a ground station in view. That cuts against the disruption framing. Putting the base station on the satellite implements more of the standardized radio access network, so the licensing surface grows as the architecture advances toward what 6G makes the default.

The competitive threat is real and the standards point does not soften it. SpaceX can undercut carrier pricing, and it can buy a carrier if it wants the dense urban capacity that satellites handle poorly. Grant all of it. None of it changes who created the interoperability. Engineers across the contributing firms built the air interface that lets a phone in a dead zone find a satellite, and they hold patents on it. Capital buys spectrum and launch vehicles. It does not buy the standard, and the standard is licensed. Even the spectrum is standards-bound. SpaceX is aligning the EchoStar bands to 3GPP band designations such as Band n70, already supported in handsets, so ordinary phones recognize them. Spectrum that maps to no standardized band the handsets implement is inert.

That makes SpaceX an implementer, and only an implementer. It contributes nothing to the cellular standards and declares no SEPs, so it carries no licensing commitment of its own. The firms that built the standard hold those obligations. SpaceX’s duty runs the other way, to take a license from them. Musk calls patents “for the weak” and says they mostly block rivals rather than advance anything, and SpaceX patents almost nothing. SpaceX can pay. The open question is whether it will license in good faith, on Fair, Reasonable and Non-Discriminatory (”FRAND”) terms, for technology it did not invent.

6G makes the satellite layer native, not a bolt-on

Release 17 added NTN late, and Releases 18 and 19 extended it onto a design built for towers first. 3GPP’s Release 21, the first 6G release, runs to a functional freeze in 2027 and a final code freeze in 2029. Boston Consulting Group and Qualcomm, in their February 2026 report, describe 6G as a layered architecture that unifies terrestrial and non-terrestrial layers, with satellites complementing higher-band access for global reach. The same report estimates that 5G has already delivered over $1 trillion in global economic impact, projects approximately $6.1 trillion by 2030 and approximately $18.2 trillion by 2035, and dates 6G to precommercial launch in 2028 and commercial launch in 2029. (link)

5G NTN (today) 6G NTN (coming) Origin Added in Release 17 (2022) Designed in from Release 21 Architecture Satellite bolted onto a terrestrial-first system Terrestrial and non-terrestrial unified Service in market Messaging and low-rate data Integrated broadband and sensing (target) Licensing surface NTN SEPs emerging NTN and Direct-to-Device SEPs central

The trajectory runs toward convergence, not toward satellites replacing cellular. A converged network leans on licensed SEPs harder, and brings more vendors onto a common standard. The satellite layer that looks proprietary in 2026 becomes a standardized layer of the 6G system by the end of the decade.

Upgrade velocity is the edge the takeover math misses

The fourth-carrier framing measures Starlink against the incumbents on their own terms: towers, spectrum, subscribers, the price of buying T-Mobile. The larger move is how fast Starlink can roll the network forward. The onboard base stations are software-defined, so the deployed fleet can take new 3GPP releases and features through updates, within the limits of what the radio hardware supports. Low Earth orbit satellites last about five years, so SpaceX refreshes the hardware continuously through replenishment launches rather than on a slow tower-swap cycle. A carrier upgrading to 6G replaces radios across hundreds of thousands of cell sites. Starlink pushes code to the constellation and replaces the constellation every five years. That integrated, proprietary model is the opposite of the industry’s Open Radio Access Network (”Open RAN”) bet, which wagered that open interfaces and disaggregated software would loosen the integrated vendors' grip and instead reconcentrated on them, with Huawei, Ericsson and Nokia holding about three-quarters of the market by Omdia's 2024 count. The genuine disruptor won with integration and a licensed standard, not open software.

The limits are real, and the honest version of the claim respects them. Satellites cannot match terrestrial capacity and density in cities, where most revenue sits, and spectrum authorization runs country by country. “Supplant the carriers” overstates the dense-urban case. The defensible claim is that Starlink becomes the global coverage and roaming layer, competes hard in rural, maritime, aviation and vehicle markets, and pressures carrier economics from the edge inward, while dense urban capacity stays terrestrial for now.

Upgrade velocity points one direction for licensing. A software-defined, continuously replenished network implements a widening set of standardized features across its life and across generations. The relationship with the standard is ongoing, not a one-time hardware purchase, and it expands every time the fleet adopts a new release. That sharpens the willing-licensee question for an operator whose marginal cost of adding standardized capability is a software push. The cheaper it is to implement more of the standard, the stronger the incentive to under-declare what you owe for it.

The spectrum map undercuts “true global coverage”

The 6 GHz band runs 1200 MHz, from 5.925 to 7.125 GHz. The FCC opened the entire band for unlicensed use in 2020, the foundation of Wi-Fi 6E and Wi-Fi 7. China reserved the entire band for licensed International Mobile Telecommunications (”IMT”) in 2023, for 5G and 6G. Europe, the United Kingdom and India released the lower portion for Wi-Fi and left the upper portion open. WRC-27 will decide the upper band.

Keep two spectrum stories apart. The 65 MHz Starlink bought from EchoStar sits around 2 GHz, a low-mid-band layer prized for coverage and indoor reach and already mapped to 3GPP handset bands. The 6 GHz and 7 GHz contest is a separate, higher-band fight over 6G capacity. Starlink’s near-term service rides the first. The 6G race turns on the second.

The US bet has paid off for Wi-Fi. Unlicensed 6 GHz carries large consumer and enterprise value, and that case stands on its own. The US has also entrenched it. In January 2026 the FCC expanded unlicensed 6 GHz use further, and the Winning the 6G Race memorandum signals no appetite to reallocate any of the band. The cost shows up elsewhere. China harmonized 1200 MHz of mid-band for licensed mobile and moved first on the upper 6 GHz for IMT. The US kept all 1200 MHz of 6 GHz unlicensed and is pursuing its 6G licensed spectrum in the adjacent 7 GHz band (7.125 to 7.4 GHz) and other mid-bands, which leaves US operators without the harmonized 6 GHz footprint China is building on. IEEE Spectrum frames the result as the three largest economic blocs drifting toward incompatible spectrum models. “Seamless global coverage” runs into national spectrum rights and diverging band plans. The contest over the upper band is a contest over whose wireless architecture the rest of the world adopts.

What this means for 2028

NTIA launched Mission LA 2028 to stage industry-led demonstrations of early 6G at the Summer Olympics, and received signed letters of intent from interested entities. (link) T-Mobile, the Games’ official telecommunications provider, will run 5G-Advanced with network slicing and satellite coverage across more than 100 venues. The Presidential Memorandum “Winning the 6G Race” names mid-band spectrum as the priority and directs NTIA to study additional bands, including portions of 7.125 to 7.4 GHz. (link) WRC-27 and the 2028 Games sit as checkpoints on the path to 6G standards in 2029.

A second pressure runs alongside the licensing question. The America’s AI Action Plan promotes open source as a path to innovation, and ETSI has begun examining how to make its specification process more amenable to royalty-free and open-source contributions. (link) Open source builds real value, and the case for it in many settings is strong. The strategic cost lands when it becomes the default. A US firm that spends billions to develop a model or a standardized technology, then releases it royalty-free, hands the result to a rival that did not fund it. China runs that play at scale. Chinese open-weight AI models now lead the open-source field. Alibaba’s Qwen overtook Meta’s Llama in October 2025 as the most-downloaded open model family on Hugging Face, and by May 2026 eleven of the twenty most-downloaded text-generation models on the platform were Qwen variants, against three from Llama. (link) A majority of the tokens routed through OpenRouter, a large neutral model router, now run on Chinese-built models. China pairs that engineering output with China Standards 2035, a State Council strategy that coordinates standards work in 5G, artificial intelligence and other critical technologies and pushes Chinese participation in the international bodies. (link) Privileging royalty-free models in the name of transparency or competition would weaken the business-model competition that pays for the next generation of standardized R&D, which is the same R&D the 6G race depends on.

The demonstrations are real and the ambition is correct. The risk is internal. A 6G strategy cannot run on capital and spectrum alone. It runs on the standardized R&D that contributors fund and license, and that funding depends on enforceable SEP rights. The Department of Justice’s recent statements of interest, which protect SEP holders’ ability to license and enforce, point the right way. A return to the posture that treated SEP holders as the competition problem would defund the pipeline the Olympics are meant to showcase. Two priorities follow for policymakers heading into 2028: protect the licensing and enforcement architecture that pays for standardized innovation, and close the mid-band spectrum gap before WRC-27 hardens the divide.

The disruption story will keep running, because a rocket company entering wireless is a good story. Whoever sets the standard and the spectrum map writes the licensing rules that SpaceX and the carriers will both operate inside. That is where the outcome is decided, and it draws far less coverage than the takeover math.

Harlan Strategies, LLC advises licensors and innovators on SEP portfolio strategy and FRAND policy, and publishes the Standards at Risk review. For a discussion of how these developments bear on a specific portfolio or policy position, contact jim@thestandardstrategy.com.

Sources

- Boston Consulting Group and Qualcomm, 6G: The Network for the Future of AI and Immersive Connectivity, Boston Consulting Group, 4, 11, 19 (2026).

- 3GPP, Solutions for NR to Support Non-Terrestrial Networks, Technical Report 38.821, Release 17 (2022).

- 3GPP, Release 21 Work Plan and Timeline, Radio Access Network plenary (2026).

- IEEE Spectrum, 6G and Wi-Fi in a Tug-of-War Over the 6 GHz Band (2026); Federal Communications Commission, Report and Order on unlicensed use of the 6 GHz band (2020), and Fourth Report and Order, ET Docket No. 18-295, FCC 26-1 (2026), expanding unlicensed 6 GHz operations; China Ministry of Industry and Information Technology, allocation of 6 GHz for IMT (2023).

- NTIA, Mission LA 2028 Letters of Intent (2026); The White House, *Winning the 6G Race*, Presidential Memorandum.

- Reuters, reporting on Financial Times coverage of SpaceX Starlink mobile service and the EchoStar spectrum acquisition (2026); TD Cowen (G. Williams) and Oppenheimer analyst notes.

- The White House, Winning the AI Race: America’s AI Action Plan (2025); ETSI IPR Special Committee, work on the treatment of open-source and royalty-free contributions (2026).

- On the Chinese open-weight lead: Xinhua and Alibaba Qwen team announcements (January 2026); Hugging Face Hub API snapshot, 14 May 2026 (eleven of the top twenty text-generation models are Qwen variants, roughly 100 million combined downloads, against roughly 20 million for the Llama models); OpenRouter token-routing analysis showing a majority, approximately 61 percent, on Chinese-built models (2026).

- Central Committee of the Communist Party of China and the State Council, National Standardization Development Outline (2021), and the National Standardization Development Action Plans (2022, 2024), the operative documents of the China Standards 2035 strategy.

- Federal Communications Commission, approval of the EchoStar to SpaceX spectrum assignments (AWS-3, AWS-4 and H-Block), with technology-neutral and buildout waivers (2026); on band alignment, analysis of SpaceX’s mapping of the EchoStar holdings to 3GPP band designations including Band n70.

- On Starlink’s approach and architecture: the constellation runs conventional LTE base stations with proprietary adaptations for the satellite link, a non-3GPP-NTN-compliant design rather than the Release 17 NR-NTN profiles, per a crowdsourced-measurement study of Starlink Direct to Cell (arXiv, 2025); on transparent (bent-pipe) versus regenerative payloads under 3GPP, the onboard eNodeB and Gen2 inter-satellite laser links, and the roughly five-year operational life of low Earth orbit satellites with continuous fleet replenishment.

- Musk, remarks on Jay Leno’s Garage (CNBC, 2022), calling patents “for the weak” and a blocking technique; and Wired (2012), on SpaceX’s decision not to patent given Chinese competition.

- On Open RAN concentration: Omdia data showing the combined RAN market share of Huawei, Ericsson and Nokia rising to roughly 77.5 percent in 2024, and Dell’Oro analysis (S. Pongratz) finding supplier diversity fading across most regions; AT&T’s $14 billion Ericsson agreement as an effectively single-vendor deployment; Dish/EchoStar and Rakuten greenfield losses; industry reporting through Mobile World Congress 2026.

Disclosure: I was employed by PanOptis during 2014 and 2015, roughly a decade before this appeal and two years before PanOptis first reached out to Apple. That employment is recorded on my public LinkedIn profile. I did not include it in the original intervention application, but following the observation from Apple’s outside counsel, Wilmer Hale, I promptly amended said application.

TL;DR

• Courts deciding Fair, Reasonable and Non-Discriminatory (”FRAND”) cases reach for the willing-licensor/willing-licensee construct without first reading the instrument that creates the obligation. That instrument is the ETSI Intellectual Property Rights (”IPR”) Declaration, and it is governed by French law.

• Clause 6.1’s “prepared to grant” is operative language. Under French law the Declaration is an irrevocable unilateral undertaking (engagement unilatéral de volonté). The phrase records a present, continuously maintained state of readiness, backed by years of research, prosecution, and essentiality work. It carries specific legal weight and is not boilerplate.

• Readiness runs both ways. The SEP owner’s undertaking does not evaporate, but the equitable and contractual protections the framework extends to an implementer are not unconditional. An implementer shown to be unwilling cannot claim them in full.

• For Optis v Apple, that reading bears directly on the comparables methodology behind Grounds 1 and 2, and on the limitation-periods question in Ground 4.

I. What kind of institution ETSI is

This is an analysis of Standard-Essential Patent (”SEP”) licensing and the construction of one obligation, “prepared to grant.” It starts where the case law seldom does: with the institution that produced the technology and the document that records the licensing commitment.

Distinguish Standards Development Organizations (”SDOs”) from Standards-Setting Organizations (”SSOs”). SSOs such as ANSI, BSI, or ISO set standards in the conventional sense, prescribing specifications such as paper dimensions, screw thread pitches, or safety classifications. And SDOs, such as The European Telecommunications Standards Institute (”ETSI”) and the International Telecommunication Union (”ITU”), are a fundamentally different kind of institution. They do not simply codify existing practice. They facilitate an iterative, consensus-based technical development process among a wide range of stakeholders, including network operators, device manufacturers, chipset suppliers, research institutions, and patent holders, whose commercial interests are often in direct and sometimes acute conflict with one another.

The technologies that emerge from this process are not “inserted” by participants with issued patents to promote. They are the product of sustained, contested, multi-party negotiation in which any given technical solution reflects compromise among competing approaches, some of which happen to be covered by patents that participants have declared as potentially essential.

II. Begin with the text, and the law that governs it

Elementary legal method requires that analysis of any contractual obligation begin with the text of the instrument that creates it, construed in accordance with its governing law. For the ETSI IPR Declaration, those requirements point in the same direction, and yet courts routinely bypass them, proceeding directly to policy-level reasoning about the purposes of FRAND without first reading the document that gives rise to the obligation they purport to apply.

ETSI is headquartered in Sophia Antipolis, France. The ETSI IPR Policy, which forms part of the ETSI Rules of Procedure, expressly provides that it is governed by the laws of France. That is not incidental. Under French law, the ETSI IPR Declaration is properly characterized as an irrevocable unilateral undertaking (engagement unilatéral de volonté), by which the declarant binds itself, in advance and without condition, to make licenses available to all implementers on FRAND terms. This is a distinct and well-recognized instrument under French civil law. It creates an obligation on the declarant that is enforceable by third-party beneficiaries, namely implementers, without the need for a bilateral agreement to have been concluded. The consequences of that characterization have profound implications for how the scope and limits of the FRAND obligation should be understood, implications that no court applying English or United States law has engaged with in any systematic way.

The text of the Declaration must be the starting point. The ETSI IPR Declaration form requires the declarant to confirm, in the exact language of Clause 6.1, that it is “prepared to grant irrevocable licences on fair, reasonable and non-discriminatory (’FRAND’) terms and conditions.” (emphasis added) Those three words, “prepared to grant,” are the operative language of the obligation. They are not preamble. They are not a mere statement of intent. On a plain reading they are the legal act the Declaration performs: the declarant represents, as a matter of binding commitment, that it is in a state of readiness to grant a license. That representation has a specific legal character. It is not an offer in the contractual sense. It is not a promise to make an offer in future. It is a present-tense declaration of preparedness that creates an ongoing obligation to remain in that posture.

What courts consistently fail to appreciate is that this state of readiness is neither passive nor costless, and it is not a single event. It is the expression of a substantial, multi-stage technical and legal undertaking that both precedes and continues after the filing of the Declaration, and that gives the “prepared to grant” commitment its substantive content. An ETSI IPR Declaration is not a confirmation of established essentiality. The Policy and the form require the declarant to act on a present good-faith belief: the form states that it is the “declarant’s present belief that the IPR(s) disclosed may be or may become ESSENTIAL” to the relevant standard. That formulation is deliberately prospective. It covers pending patent applications as well as granted patents. It acknowledges that the standard may still be evolving at the time of declaration, and that the relationship between the declared claims and the final standard text is not yet fully determined.

To reach the point of filing, the declarant will have invested in research and development, participated actively in the contested working group process, and identified patent claims it believes are or may become essential to the normative requirements of the emerging standard. The declarant must then map those claims with care against the adopted technical specification, because a claim is only essential if it is impossible to implement the relevant part of the standard without infringing it, with no technical workaround available. That essentiality analysis is itself a demanding exercise of legal and technical judgment, requiring a granular comparison of claim language against mandatory provisions of standards documents that may run to thousands of pages.

The declarant must also prosecute its patent applications through one or more national or regional patent offices. That prosecution is not a formality. It is an interactive, adversarial process in which claims are examined, objections are raised, and amendments are made. It must track the evolution of the standard itself, because the standard may be modified during the examination period, and claims drafted to capture an earlier version of the specification may need revision to maintain their essential character relative to the final adopted text. Amendments required by the patent office may alter the scope of the claims in ways that affect their relationship to the standard. Changes made to the standard during the 3GPP release process may affect whether previously drafted claims remain essential to the final specification. Managing that bidirectional interaction between patent prosecution and a living standards process is a sophisticated, resource-intensive undertaking that is invisible to popular discourse about SEP licensing, and almost equally invisible in the case law.

The filing of the Declaration, and the representation that the declarant is “prepared to grant,” is therefore not the endpoint of a completed investment cycle. It is a binding, point-in-time commitment made within an ongoing and dynamic process. The ETSI system reflects this. Declarants are encouraged to update their declarations as circumstances develop. A declarant may formally notify the ETSI Secretariat that a patent it had previously declared as potentially essential is no longer considered essential, and that update is recorded in a dedicated column of the ETSI IPR online database, which tracks the full modification history of each declaration. The mechanism is not theoretical. Parties have used it where prosecution outcomes, standard amendments, or further essentiality analysis showed that a declared patent does not in fact cover the normative requirements without an available workaround. Its existence and use make clear that “prepared to grant” is a continuously maintained state of readiness, subject to ongoing legal and technical evaluation against the evolving relationship between the declared claims and the adopted standard text. Courts that treat the Declaration as a threshold to a royalty negotiation, and read “prepared to grant” as boilerplate, have misunderstood what the Declaration represents, why ETSI designed it to be a living and updatable commitment, and why its language carries the legal weight it does under French law.

III. Readiness runs both ways

A further and significant consequence flows from the dynamic and bilateral character of the FRAND undertaking. Where an implementer has been found by a court, or where its conduct plainly demonstrates, that it is not prepared to take a license on FRAND terms, a serious question arises as to whether the equitable and contractual protections that the FRAND framework extends to implementers remain available to it. This is not to say that the SEP owner’s irrevocable undertaking under French law simply evaporates: the “prepared to grant” commitment, once made, binds the declarant. But the FRAND framework is not a one-sided instrument of protection for implementers. It is a bilateral framework of mutual obligations. The framework confers benefits on the implementer precisely because and to the extent that the implementer occupies the corresponding bilateral position: that of a party who is, and who conducts itself as, prepared to take. Where that posture is absent, where the implementer has been declared or has demonstrated through its conduct that it is unwilling to accept FRAND terms, the contractual and equitable logic of the bilateral framework is disrupted at its foundation.

The UK Supreme Court has already recognized, in the context of injunctive relief, that an implementer who refuses a FRAND offer forfeits the protection against injunction that the framework would otherwise provide. The Court of Justice of the European Union (”CJEU”) in Huawei v ZTE similarly structured the framework as a sequence of bilateral obligations with consequences for the implementer who fails to respond as a willing licensee.[1] The argument advanced here is the natural extension of that principle: where an implementer is not, and has not conducted itself as, prepared to take, it cannot fully claim the benefit of a framework whose protections are premised on bilateral preparedness. Courts conducting a FRAND rate determination should not proceed as though they are dealing with a willing licensee in the full bilateral sense that the ETSI IPR Policy contemplates, if the evidence does not support that characterization of the implementer before them.

The significance of “prepared to grant,” correctly understood, is that it creates a reciprocal legal framework. The declarant’s preparedness to grant is the necessary correlative of the implementer’s entitlement to receive. The relationship is structurally analogous to an invitation to treat: it establishes the conditions within which a binding transaction can take place, and it places both parties under corresponding obligations as to their posture and conduct. An invitation to treat is legally meaningless if no party is in a position to accept. The “prepared to grant” commitment is commercially meaningless if there is no willing licensee prepared to take the license the SEP owner is prepared to grant. The willing licensor and the willing licensee are not independent constructs that a court can analyze in isolation. They are two sides of a single bilateral framework that the ETSI IPR Policy deliberately created.

Courts repeatedly commit the opposite error. They ask what a willing licensee must pay, or what rate a court should impose, without first asking whether the party claiming the benefit of the framework has itself discharged the obligation the framework places on it. They treat “prepared to grant” as throat-clearing before the real analysis begins. It is the foundation of the legal structure. Legal Scholar Adam Mossoff has shown, through systematic analysis of the Policy text, its drafting history, and the failed attempts to amend it, that the FRAND commitment was deliberately structured to preserve the SEP owner’s legal remedies while imposing an affirmative and unconditional obligation of readiness to license.[2] A court that begins anywhere other than the text of the Declaration, read under French law, has already made an error of method.

IV. The virtuous cycle, and what it means for rate-setting

How the technical content of a standard such as 4G LTE comes into existence matters here. Participants in ETSI working groups do not arrive and propose to insert their patented technologies into the standard. Technical proposals are submitted, debated, tested against rival approaches, refined through successive drafts, and resolved through consensus among participants whose interests are often sharply opposed. A manufacturer and a chipset supplier may hold directly conflicting preferences on a given question. An operator may have priorities that diverge from both. A research institution may advocate a theoretically superior approach that industry regards as impractical at scale. Out of that conflict, a technical solution emerges. In some cases, patents that participants declared as potentially essential turn out to cover aspects of the eventual solution. In many cases they do not. The ETSI IPR Policy exists to manage that uncertainty, committing participants in advance to license whatever of their declared-essential patents actually are incorporated into adopted standards. That commitment is the “prepared to grant” obligation. It is the price of participation in a process that produces global interoperability.

Participation requires SEP holders to invest considerable resources and time in research and development, to develop new technology, to patent it worldwide, and to contribute it for potential inclusion in a standard, all without any guarantee that their patents will be incorporated, or that the standard will be accepted by the market.[3] The Policy recognizes this reality. Its design reflects an understanding that technology innovators must have the opportunity to realize a proper return on their investment, balanced against affording implementers access to the technology so they can develop new products and services.

That balance operates through what Taffet describes as a virtuous cycle of invention, implementation, and consumption. The SEP holder invests in research and development, contributes the resulting technology to the standards process, realizes a return through licensing, and reinvests in further innovation, which contributes to the next generation of standards. As a repeat player, the SEP holder has strong incentives to make its technology broadly available rather than to restrict access: wider adoption increases the number of licensees, reduces marginal costs, and grows licensing revenues. The long-term business model depends on the cycle continuing. A FRAND determination that suppresses royalties below the level genuine bilateral negotiation would produce disrupts the cycle at exactly the point where it depends on the return being realized.[3] [4]

Those who argue that the Court of Appeal’s rate determination was too high tend to characterize SEP licensing as a source of rent extraction and strategic opportunism. That characterization is not supported by the empirical record. More than two decades of regulatory scrutiny and academic analysis have failed to produce evidence that SEP licensing has systematically harmed innovation, deterred standards participation, or caused the predicted parade of horribles.[3] What the empirical record does show is that the open, voluntary, consensus-based, private-sector-led model of standards development that ETSI exemplifies has produced continuous technical improvements, new industries, new products and services, and consumer benefits worldwide. The legal framework that sustains that record is the one the ETSI IPR Policy created, centered on the “prepared to grant” obligation and the bilateral framework of obligations it establishes.

V. Why this matters for Optis v Apple

Apple’s Grounds of Appeal advance five challenges to the Court of Appeal’s judgment in Optis v Apple.[5] What is conspicuously absent from all five Grounds, and from the submissions of the other interveners who have filed in support of Apple, is any engagement with the foundational question of what it means, as a matter of the text of the ETSI IPR Declaration construed under its governing French law, for an SEP owner to have arrived at a genuine state of readiness to grant, and for an implementer to be correspondingly obligated to be prepared to take.

First, the “prepared to grant” obligation establishes the normative baseline from which FRAND rate-setting proceeds. The Policy does not start from the implementer’s preference. It starts from the SEP owner’s declared commitment, grounded in the upstream investment cycle of research, prosecution, and essentiality mapping. A court determining what a hypothetical willing licensor and willing licensee would agree must take as its starting point that the licensor has already committed, contractually and irrevocably under French law, to offer a license, and that this commitment reflects a substantial prior investment the FRAND rate must be adequate to sustain. Courts that collapse this into a generalized inquiry into what the market would bear, without anchoring the analysis in the full legal and economic content of the “prepared to grant” posture, misread the regime they are applying.

Second, and with direct relevance to the comparables methodology at the heart of Grounds 1 and 2, the upstream investment captured in the “prepared to grant” declaration also bears on how portfolios should be valued relative to one another. Not all ETSI IPR Declarations represent equivalent investments. A portfolio assembled through sustained original R&D, active contribution to the SDO working group process, and careful prosecution of claims tracked against the evolving standard represents a materially different kind of commitment from one assembled through acquisition of patents whose essentiality was established, and whose prosecution was managed, by others. The Court of Appeal’s observation that Optis’s portfolio was at best average in quality does not alter the fact that its patents were declared and prosecuted through the ETSI process, and that the “prepared to grant” commitment those declarations represent is the legally operative fact for rate-setting purposes. What the comparables analysis must avoid is treating licenses to portfolios representing fundamentally different levels of upstream investment as interchangeable data points. Licenses extracted through hold-out, compressing rates below what a willing licensee genuinely prepared to honor its side of the ETSI bargain would have agreed, compound that error further.

Third, the "prepared to grant" framework provides the correct lens for Ground 4, the limitation-periods question. The ETSI Policy imposes no obligations on non-member implementers. The expectation that an implementer act rather than wait comes from the FRAND case law. ETSI maintains a public IPR database of declared-essential patents, and its accompanying notice directs a potential licensee to contact the declarant before implementing a declared SEP. [6] Any implementer can therefore identify the patent holder. Huawei v ZTE then requires an implementer that keeps using the patents while terms are negotiated to provide security for its past and ongoing use.[1] An implementer that deploys the standard without approaching the declarants it can identify, and relies on delay to erode its liability, is not conducting itself as a willing licensee. The Court of Appeal's conclusion in InterDigital v Lenovo,[7] upheld here, that royalties run from the date of first infringement follows from that conduct standard. Ground 4 should fail.

VI. The broader stakes

For close to twenty years I have observed that the “prepared to grant” language of Clause 6.1 is almost never the subject of direct judicial analysis, that the governing law of the Declaration is almost never engaged with as the primary source of interpretive authority, and that the upstream investment cycle which gives the commitment its substantive content is almost entirely invisible in the case law. Courts reach for the FRAND obligation as an all-purpose framework without first asking what the text of the Declaration actually requires, of whom it requires it, what that requirement presupposes about the prior work of the declarant, and what French law says about the character and consequences of the obligation. The result is a body of case law less principled, less consistent, and more vulnerable to strategic manipulation than it should be.

The Supreme Court has the opportunity to correct that. Unwired Planet v Huawei[8] established the global FRAND jurisdiction and the high-level framework. What remains underdeveloped, in this jurisdiction and internationally, is a rigorous account grounded in the text and governing law of the Declaration: what “prepared to grant” means, what upstream investment it presupposes, what bilateral consequences flow from it, how it connects to the economic structure of the virtuous innovation cycle, and how all of that should shape both the willing-licensor/willing-licensee construct and the evaluation of comparables.

The standards ecosystem that underpins global cellular communications, and increasingly the Internet of Things (”IoT”), connected vehicles, healthcare technology, and industrial connectivity, depends on a genuine, functioning bilateral commitment of the kind the ETSI IPR Policy created. SEP owners must be prepared to grant, having done the work such preparedness requires. Implementers must be prepared to take, having accepted that the cost of access to globally interoperable standardized technology is a commercially reasonable royalty paid to those who made it possible. That framework of mutual obligations is what makes the iterative, consensus-based standards process economically viable for the upstream innovators whose sustained investment and active participation sustain it. A judgment that gives proper legal weight to both sides of that framework, beginning from the text of the instrument that created it and the law that governs it, would provide the clarity and predictability the global technology licensing ecosystem urgently needs.

[1]: Huawei Technologies Co Ltd v ZTE Corp, Case C-170/13, [2015] 5 C.M.L.R. 14 (CJEU).

[2]: Mossoff, Adam, Patent Injunctions and the FRAND Commitment: A Case Study in the ETSI Intellectual Property Rights Policy, Berkeley Technology Law Journal, Vol. 38, p. 487 (2023).

[3]: Taffet, Richard S., Standards Development and Innovation: Time to Stop Talking About SEP Licensing, Competition Policy International TechREG Chronicle (2025).

[4]: Dasgupta, Kalyan & Teece, David J., Protecting Innovation in the Mobile Wireless Ecosystem: Understanding and Addressing “Hold-Out”, Berkeley Technology Law Journal (2023).

[5]: Optis Cellular Technology LLC v Apple Retail UK Ltd, [2025] EWCA Civ 553 (the unredacted judgment; the public redacted version is [2025] EWCA Civ 552).

[6]: ETSI, Intellectual Property Rights (IPRs), ETSI IPR Online Database (a potential licensee should contact the declarant before implementing a declared SEP); see also ETSI Special Report SR 000 314.

[7]: InterDigital Technology Corp v Lenovo Group Ltd, [2024] EWCA Civ 743.

[8]: Unwired Planet International Ltd v Huawei Technologies Co Ltd, [2020] UKSC 37.

Harlan Strategies, LLC advises licensors and innovators on SEP portfolio strategy and FRAND policy, and publishes the Standards at Risk. For a discussion of how these developments bear on a specific portfolio or policy position, contact jim@thestandardstrategy.com.

The comment period closed May 21, 2026. I filed. Then I got curious about who else did.

Part One: What I Filed, and Why

The press release asked three questions. What topics would benefit from additional guidance, including joint licensing arrangements? What new technologies and business models warrant clarity? What significant legal, economic, or technological developments should inform any revision?

All three questions converge on the same domain: voluntary standards development and the standards-essential patent (”SEP”) licensing framework that makes it commercially viable. That convergence is not accidental. The standards ecosystem sits at the intersection of competitor collaboration, intellectual property licensing, and technology deployment at scale. It is precisely where updated antitrust guidance will have the most consequential effects.

My comment made six arguments.

The first is foundational. Voluntary, private-sector standards development through bodies such as the Third Generation Partnership Project (”3GPP”), the Institute of Electrical and Electronics Engineers (”IEEE”), and the European Telecommunications Standards Institute (”ETSI”) is among the most pro-competitive forms of competitor collaboration in the modern economy. The 3GPP process has, over more than 25 years, brought hundreds of competing firms together to reach technical consensus on 3G, 4G, 5G, and now 6G. Boston Consulting Group data show that 5G-enabled applications have already generated over $1 trillion in global economic impact, with cumulative value projected at $18.2 trillion by 2035. These outcomes did not happen despite competitor collaboration. They happened because of it. Any guidance that treats standards participation as presumptively suspect will erode the framework that produced them.

The FRAND declaration is an irrevocable unilateral undertaking under French law, the governing law of the ETSI intellectual property rights policy. A patent holder who declares under ETSI's policy is "prepared to grant" a license on FRAND terms. That language is operative, not prefatory. It records a continuously maintained state of readiness. It does not compel a grant, and it does not authorize an implementer to deploy a patented technology without a negotiated license. The bilateral framework of mutual obligations runs in both directions: the declarant commits to license on reasonable terms, and the implementer commits to negotiate in good faith and take a license before deploying. Guidance that treats the FRAND commitment as primarily an access guarantee, and ignores the reciprocal obligation on implementers, misreads the framework and will produce exactly the holdout dynamic the empirical record has documented.

The third argument is the one that regulatory commentary has most consistently underdeveloped: implementer holdout. For two decades, the dominant concern in policy discussions has been patent hold-up by SEP holders. The empirical record has not confirmed it. Bowman Heiden and Justus Baron, writing in the Harvard Journal of Law and Technology in 2024, found the opposite: a royalty gap of $7 to $28 billion annually as of 2021, representing systematic undercompensation of SEP holders attributable in significant part to implementer holdout. Anne Layne-Farrar’s 2014 survey of 15 years of economic literature for the OECD reached the same conclusion from a different angle: empirical studies had not shown hold-up to be a common problem. Kirti Gupta and Urska Petrovčič documented systemic holdout behavior through comprehensive court case analysis published in the Berkeley Technology Law Journal. Guidance premised on a theory the data have not confirmed will produce over-deterrence without any offsetting benefit.

The fourth argument draws on the IEEE’s experience after the DOJ’s 2015 Business Review Letter as a natural experiment. Following the BRL’s endorsement of the revised IEEE intellectual property rights policy, new positive letters of assurance for 802.11 Wi-Fi standards dropped 90 percent. Negative letters reached an all-time high in 2016. In September 2022, IEEE rescinded the core provisions, including those restricting injunctive relief. The sequence is about as close to a controlled test of the interventionist hypothesis as the standards policy record provides. The result was unambiguous. Updated guidance should account for it.

The fifth argument addresses joint licensing and patent pools specifically, because the agencies called them out by name as a topic warranting guidance. Well-designed patent pools and platforms that aggregate SEP holders and offer portfolio licenses reduce transaction costs, increase access for smaller implementers, and facilitate licensing where portfolios are fragmented. They have historically been recognized as pro-competitive. Updated guidance should reaffirm that recognition and address concerns through transparent governance requirements rather than presumptive suspicion.

The sixth argument connects the guidance question to national technology strategy. President Trump stated in December 2025 that 6G will be foundational to U.S. national security and pivotal to the development of artificial intelligence. Senate Commerce Committee Chairman Ted Cruz has warned that failure to lead in 6G would allow Huawei to create the backbone, handicapping U.S. efforts in AI, quantum, and semiconductors. BCG documents that China, South Korea, and India have each articulated explicit national targets for 6G standard-related patent shares. Guidance that raises the cost of enforcing SEP rights will disproportionately disadvantage the smaller innovators and startups whose R&D intensity makes standards leadership possible in the first place. Standards leadership and IP enforcement are not in tension. They are complementary elements of the same policy framework.

The comment closed with a structural point. The December 2024 withdrawal left businesses engaged in standards development and SEP licensing without any surviving framework document. The 2000 guidelines are gone. The 1995 IP licensing guidelines on which they relied are also withdrawn. The inquiry is the opportunity to fill that gap with guidance grounded in what the empirical record actually shows, rather than two decades of theoretical concerns the data have not confirmed.

Part Two: Who Else Filed, and How I Found Out

After the deadline passed, I wanted to see the full comment record. The answer is technically public. Getting to it is another matter.

Regulations.gov lists comments individually. There is no bulk export, no docket-level CSV, and no obvious path to a structured index of who filed and where their submissions live. For a practitioner who wants to survey the record before the agencies act on it, the portal is nominally open and practically clunky.

A short Python script changes that.

The Regulations.gov REST API v4 is free to access with an api.data.gov key, passed via an X-Api-Key header and kept in a .env file. The architecture has one non-obvious wrinkle that cost some time to work out. The /comments endpoint has no docket filter. You cannot query directly by docket ID. Instead, you fetch the docket’s documents first, read each document’s internal objectId, then query comments filtered by that ID, paginating through the results. For each comment in the list view, a second detail call is required to retrieve the submitter’s name and any attachment PDF links, because those fields do not appear in the list response.

The output is a CSV: one row per comment, with submitter name, comment ID, a direct Regulations.gov link to the comment, and a direct link to any PDF attachment. For docket ATR-2026-0001, that produced 100 comments and 107 rows. A handful of filers submitted multiple PDFs.

From there, a second script generates a landscape PDF of the same table, with clickable links, a repeating header row, and a translucent diagonal watermark. That PDF is what I attached to the LinkedIn post that preceded this article.

The build took a single Claude Code session, including debugging a transient API error on the detail call loop and getting the watermark opacity to render correctly. The architectural decisions, including the scope choice to index the record without downloading or summarizing any filing, were deliberate human calls made before the session started. The code executed them.

The scope choice deserves a word. The script is informational. It lists who commented and links their submissions. It does not download filings, summarize arguments, or take any position on the rulemaking. A neutral index of the public record is useful to everyone regardless of where they stand on the substance. Building editorial selection into the tool would have narrowed its utility and raised questions about what was left out. A clean index leaves those judgments to the reader.

The script is reusable. Swap the docket ID and it works for any Regulations.gov proceeding. The code is at github.com/theharlans/standards-at-risk.

A closing observation

One hundred comments on a DOJ antitrust inquiry is a real record. The agencies asked three specific questions and got substantive responses from practitioners, academics, trade associations, and firms across the technology sector. Whether the resulting guidance reflects the weight of that record, or selects from it, is the question worth watching.

My comment argued that the empirical case for treating SEP licensing as a presumptive antitrust risk has never been made, that implementer holdout is the underweighted market failure in this domain, and that guidance calibrated to theoretical harms will produce real deterrence of the R&D investment and standards collaboration the agencies say they want to encourage.

The comment is in the record. The index of who else filed is now public. The next move belongs to the agencies.

Jim Harlan is a standards and SEP licensing practitioner and the author of Standards at Risk.

What Was Announced

This morning InterDigital published the following on Globe Newswire:

“InterDigital, Inc. today announced that it has entered into a patent license agreement with Amazon, covering Amazon’s services and devices, including Amazon Prime Video. The parties have agreed to resolve all pending litigation and will enter into binding arbitration to determine the final terms of the new agreement.”

That is the Samsung January 2023 announcement structure, word for word.

• Binding arbitration to determine final terms.

• All pending litigation resolved.

• Revenue recognition begins immediately at a conservative level.

The final award comes in approximately 18 to 24 months -- call it late 2027 to mid-2028 based on the Samsung (30 months) and Lenovo (18 to 24 months) precedents.

The market confirmed the two-layer framework in real time:

IDCC surged as much as 18 percent intraday on the announcement. As of mid-morning the stock was up approximately 9.6 percent. This on a stock that had been down 29 percent over the prior six months, trading at $253.91 before the announcement.

The market did not wait for the arbitration award. It did not wait for the final streaming services rate. It moved on the announcement alone -- exactly as the Samsung January 2023 precedent predicted. The two-layer framework the series identified in Part 5 is not just analytically correct. The market price action confirms it. The announcement is the economic event. The award is the true-up.

Amazon already held a device-level license from InterDigital covering WLAN and related device technologies -- confirmed by InterDigital’s own Q4 2024 investor materials. The streaming litigation was a completely separate workstream. InterDigital launched its Video Services licensing program as a new initiative targeting streaming platforms as a distinct licensee category that had never previously been licensed. The enforcement campaign against Amazon was InterDigital asserting that streaming services require a separate license independent of any prior device license -- not a failed renegotiation of the prior deal.

Today’s announcement covers “Amazon’s services and devices, including Amazon Prime Video.” That language bundles the device component and the new streaming services license together under a single agreement. The multi-forum enforcement campaign across six jurisdictions was the mechanism that brought Amazon to the streaming licensing table. The device component was always going to renew. The structural question was always whether streaming would be in scope under a new and separate license.

It is.

InterDigital’s July 2024 Google license -- covering cellular wireless, Wi-Fi, and HEVC video on consumer devices -- is the device-only template. The Amazon agreement adds streaming services on top of that foundation.

What the Series Predicted

This series ran across five published parts from March to May 2026, using a ten-agent Python swarm to reason over an evolving fact base. The prediction was consistent across all five parts: confidential global resolution, not a public FRAND judgment. The mechanism, timing, and confidence evolved with each run.

Here is the complete prediction record:

Part 1 (March 2026): Confidential global license, Q4 2026. Confidence: 52 percent. Primary seed document: a single JUVE Patent article. Key driver: UPC penalty exposure. The prediction was directionally correct. The timing was off by one quarter -- the announcement came in Q2 2026, two quarters earlier.

Part 2 (April 2026): Corrected the Delaware companion case error from Part 1. Added the full US docket picture. Confidence: 54%. Directionally correct.

Part 3 (April 2026): Added the validity challenge layer, the ITU Common Patent Policy analysis, and the regulatory flanking analysis. Confidence: 68%. This was the high-water mark of the series and, in retrospect, the most accurate single run in terms of direction and confidence.

Part 4 (April 30, 2026): First run anchored to an SEC-filed primary source -- InterDigital’s Q1 2026 10-Q. Added the Samsung ICC award, the Dolby California DJ action, the German court trial schedule, and the AV1 correction. Confidence: 61%.

Part 5 (May 12 to May 17, 2026): Triggered by the UK Court of Appeal’s May 12 judgment in Acer v Nokia, [2026] EWCA Civ 564. Ran six separate swarm iterations incorporating the Nokia EWCA judgment, the Samsung and Lenovo two-layer announcement framework, the patent fraction stress test, InterDigital’s 2030 ARR streaming channel roadmap, the Amazon existing device license correction, and the streaming industry collective action dynamic. Confidence across the six runs: 63, 54, 61, 64, 47, and 60 percent. Final published confidence: 60% in Q3-Q4 2026. The announcement came in Q2 2026 -- one quarter earlier than the published window.

What the Series Got Right

The directional prediction held across all five parts. Confidential resolution through binding arbitration, not a public FRAND judgment. Every run reached the same conclusion.

The mechanism was precisely right. Part 5 identified the Samsung and Lenovo arbitration announcement as the correct prediction endpoint -- not a final license, but an announcement of binding arbitration followed by conservative revenue recognition. That is exactly what happened this morning.

The streaming services characterization was right. The announcement explicitly covers “Amazon’s services and devices, including Amazon Prime Video.” This confirmed the Part 5 correction that this was InterDigital’s Video Services licensing program enforcement action, not a device patent dispute. Amazon already held a device license. The new agreement covers streaming.

The Bird and Bird connection was right. The same partner who represented Nokia in the Acer v Nokia proceedings represents InterDigital in the Amazon dispute. The Nokia EWCA decision validated the arbitration playbook that InterDigital then deployed.

The streaming licensing strategy was right. The press release describes the agreement as “an important milestone in InterDigital’s longer-term goal to expand into video streaming services licensing.” That language maps directly to the streaming licensing expansion InterDigital described at its September 2024 investor day. The $485 billion addressable market thesis now has its first licensed counterparty.

The Samsung and Lenovo timeline precedent was right. The announcement came on June 11, 2026. The final arbitral award will follow in approximately 18 to 24 months. We will not know the final streaming services rate until late 2027 or mid-2028 at the earliest.

What the Series Got Wrong

The timing was off by one to two quarters. The final published prediction was Q3-Q4 2026. The announcement came in Q2 2026. The prediction window was correct in directional terms -- before the September 2026 UK trial -- but conservative on timing. The May 28, 2026 UPC Court of Appeal hearing on the AILI, identified as the key driver, may have accelerated the timeline even before that hearing concluded.

The confidence numbers were volatile across Part 5. Six runs produced confidence levels ranging from 47 to 64 percent. The volatility reflected genuine uncertainty about endpoint definition. The final published 60 percent was the right number under the correct two-layer framework, but the journey to that number was noisier than it should have been.

The device license correction came late. The earlier Parts treated this as a device patent dispute when I knew better as evidenced in the pleadings. The correction -- Amazon already holds a device license, this is a streaming services enforcement action -- only entered the analysis in Part 5. It should have been in Part 1. Earlier identification of the existing device license would have produced sharper predictions throughout.

What Comes Next: The Q2 2026 Earnings Picture

The announcement today falls in Q2 2026. InterDigital’s Q2 2026 earnings call will reflect this agreement in the same way Q1 2023 reflected the Samsung announcement.

Expect the Q2 2026 10-Q to disclose: revenue recognition from Amazon beginning in Q2 2026 at a conservative level; a statement that the arbitration award is expected to exceed the conservative estimate and require a true-up; updated commentary on the streaming licensing program noting the first major platform license is in place; and reduced litigation expense guidance given the resolution of all pending proceedings across the UK, US, Germany, UPC, and Brazil.

The conservative recognition rate will not be disclosed. The streaming services rate will not be disclosed. The total award value will not be known for approximately 18 to 24 months.

A Note on the Methodology

This series used a ten-agent Python reasoning system to analyze an evolving public record across five published parts and twelve total swarm runs. The system ran on Claude Sonnet 4.6 via the Anthropic API. Each run cost approximately $3.50 to $4.00 at current pricing. Total cost across the series: approximately $45 to $50.

The full Python implementation -- pre-processors, agent architecture, and all supporting code -- is available at github.com/theharlans/standards-at-risk.

The methodology was challenged publicly after Parts 1 and 2. On April 9, 2026, a commentator published a piece titled “Let’s not allow AI slop to pollute the SEP ecosystem” calling the analysis hallucination dressed up as reasoning and offering a public $10,000 bet that the prediction was wrong. In Part 3, several of his substantive corrections were acknowledged and incorporated. He returned in subsequent articles to note the series was still missing key points. The bet was never accepted because engagement was never the point. The analysis was. The announcement this morning speaks for itself.

What This Means for the Streaming Industry

Amazon is now the first major SVOD and AVOD platform to enter a binding patent license agreement under InterDigital’s Video Services licensing program. The arbitral tribunal will determine the streaming services rate on a confidential basis.

Every other streaming platform -- Netflix, Disney Plus, Apple TV Plus, Paramount Plus, YouTube -- is watching. InterDigital now has a licensed counterparty in its streaming services licensing program and a rate-setting process underway that will produce a confidential benchmark. Future licensing conversations with other streaming platforms will reference this agreement.

Whether the streaming licensing theory is legally sound -- whether SEP holders can extract a separate royalty from content delivery platforms independent of device licenses already paid -- will be determined by the ICC tribunal, not by a public court. That question will not have a public answer for 18 to 24 months.

The Dolby dimension:

The Amazon settlement does not resolve the parallel Dolby proceedings. In February 2026, Dolby filed IPR2026-00252 against InterDigital’s US 9,185,268, an HDR patent. In March 2026, Dolby filed two declaratory judgment actions in the Central District of California (Case Nos. 2:26-cv-02269 and 2:26-CV-02270) asking the court to confirm that Dolby’s customers -- including Disney, TCL, and Hisense -- do not infringe six of InterDigital’s HDR patents, or that US 9,185,268 is invalid. Amazon was named as one of those customers. With Amazon now licensed, the Amazon-related motivation for Dolby’s DJ actions falls away -- but the actions themselves continue. Disney, TCL, and Hisense remain unlicensed and Dolby filed on its own behalf, not Amazon’s. The Dolby IPR at the USPTO is entirely unaffected by the Amazon settlement. Patent validity is a public right question that survives any bilateral agreement.

What changes is the commercial dynamic. InterDigital now has a licensed Amazon agreement as a reference point in negotiations with every remaining target. Disney, TCL, and Hisense face a harder argument that the streaming licensing theory is commercially unproven. The Dolby DJ actions and IPR continue, but the ground shifted this morning.

The September 2026 UK FRAND trial has been vacated. The ITC investigation is terminated. The German Regional Court proceedings are withdrawn. The UPC proceedings are suspended. All of the multi-forum enforcement pressure that this series analyzed across five parts resolved this morning with a three-paragraph press release.

That is how InterDigital wins.

Standards at Risk covers SEP licensing, FRAND jurisprudence, and global standards policy. The five-part InterDigital v Amazon series, including all source code for the analytical methodology, is available at standardsatrisk.com. This post is the author’s own analysis. It is not legal or investment advice.

TL;DR

• David Rudin’s Coffee & Standards is a well-constructed field guide to standards governance and IP policy from a practitioner with more than two decades at Microsoft. It belongs on the shelf of anyone entering this practice.

• The mechanical accuracy is not the issue. The issue I have with the publication is that its framing assumptions embed an implementer-side posture that an uninitiated reader will absorb as neutral description.

• Rudin, writing from a standards governance and transactional practice rather than a patent prosecution background, opens his patent chapter with “The Monopoly Problem” and treats the patent itself as the source of the problem FRAND is designed to solve.

• Rudin describes telecom standards development as companies independently developing technology, filing patents, and then contributing those patents to standards bodies. That is not how Standards Development Organizations (”SDOs”) like ETSI or 3GPP work. The technology and the standard emerge together, developed in concert by parties with opposing economic interests.

• Treating royalty-free (”RF”) open source and Fair, Reasonable and Non-Discriminatory (”FRAND”)-bearing standards as interchangeable delivery mechanisms for interoperability is not a neutral governance observation. The Alliance for Open Media (”AOM”) and AV1 show what happens when that framing travels unchecked into real markets.

• Open source and standards differ not only in IP framework but in governance structure, development process, and patent coverage. The Open Source Initiative (”OSI”) Open Source Definition (”OSD”) criterion 7 leaves the patent coverage question genuinely unresolved. A Benevolent Dictator for Life (”BDFL”)-governed project meets none of the due process criteria that formal standards bodies are expected to satisfy.

• The governance frameworks Rudin analyzes (ANSI, WTO Technical Barriers to Trade (”TBT”), EU Regulation 1025/2012) are silent on what happens after a standard achieves widespread adoption. That is where the litigation lives.

Coffee & Standards grew out of internal training sessions David Rudin ran for attorneys at Microsoft. It shows. The book is organized, plainly written, and grounded in genuine institutional experience. Rudin’s treatments of necessary claims, patent pool antitrust questions, and multi-party negotiation dynamics are among the better short-form practitioner accounts available.

The concern here is not accuracy on mechanics. It is what the book teaches implicitly, through the assumptions embedded in its framing, that a reader without prior exposure to the licensor side of these disputes will carry forward without examining.

What the Book Gets Right

Start with what Rudin gets right in Chapter 1, because it is genuinely useful and often missed. A standard sets the floor for interoperability, not the ceiling for innovation. The standard tells you the minimum required to interoperate. It says nothing about what you are permitted to build on top. Rudin flags this early and correctly identifies it as the source of a persistent regulatory misunderstanding: procurement policies and government mandates that treat the standard as the boundary of permitted behavior rather than the baseline for competition. That error stifles the very innovation that standards are supposed to enable. It is a point worth making clearly to any practitioner advising on public procurement or regulatory compliance.

His Chapter 4 treatment of harmonized European standards deserves particular attention for practitioners advising on EU matters. When a directive sets essential requirements and the Commission issues a standardization request, a resulting harmonized standard cited in the Official Journal gives implementers a legal presumption of compliance with the underlying requirement. Compliance with the standard equals compliance with the law. That is a substantially higher legal consequence than anything attached to a U.S. voluntary consensus standard, and it bears directly on how ETSI’s standard-setting function should be understood in the context of EU regulatory enforcement.

Chapter 7 is more balanced than most practitioner guides in this space. Rudin names the implementer holdout problem directly: an implementer who delays, ignores licensing overtures, and forces the patent holder to litigate just to recover what was already owed is not behaving as a willing licensee. Courts have addressed this unevenly across jurisdictions, and the law is not settled, but Rudin presents both hold-up and holdout as real phenomena rather than treating one as established fact and the other as industry talking point. That even-handedness is the right starting position for a practitioner who needs to advise clients on either side of a licensing dispute.

On the FRAND commitment itself, his Chapter 7 formulation is correct and worth preserving: the RAND commitment is a commitment to license, not a license itself. The standards organization facilitates the commitment but does not set rates, administer licenses, or broker negotiations. Practitioners who conflate the two, treating the FRAND commitment as a rate determination or a cap on what a licensor can recover, are making a foundational error that Rudin’s framing correctly resists.

The Development Process Problem

Chapter 7 explains why RAND works for the telecom industry this way: “Companies spend years developing wireless technology, file patents on it, and then contribute it to standards bodies with the expectation that they’ll license those patents to implementers.”

That sentence describes something. It does not describe how technology development actually works inside 3GPP or ETSI.

Before turning to how telecom technology is actually developed, it is worth pausing on Rudin’s foundational premise. Writing from a standards governance and transactional practice rather than a patent prosecution background, he opens the patent chapter with “The Monopoly Problem” and treats the patent itself as the source of the problem FRAND is designed to solve. That framing has it backwards. A patent is a negative right. It is the right to exclude others from practicing the claimed invention for a limited term. It does not grant the holder the right to make, use, or sell anything. A patent on a technical approach that nobody is required to use creates no leverage over anyone. The question of what leverage, if any, a patent holder possesses is determined by whether implementers have commercially viable alternatives, not by the existence of the patent itself.

This is not an abstract reframing. The DOJ has made it an enforcement principle. The Statement of Interest filed in Disney Enterprises, Inc. v. Interdigital, Inc. (October 6, 2025) (link) established that market power cannot be presumed simply because a patent has been incorporated into a technical standard. The Statement of Interest in Samsung Electronics Co., Ltd. v. Netlist, Inc. (April 7, 2026) (link) reinforced this, holding that a complaint cannot sufficiently allege market power based on Standard Essential Patents (”SEPs”) without affirmative assessment of alternatives to the standard and the contractual obligations the patent holder has assumed under the SDO’s patent policy. Under the rule of reason framework confirmed in Texaco Inc. v. Dagher, 547 U.S. 1 (2006) (link), market power is a threshold element requiring affirmative proof, not a presumption that follows from standardization. A rate-setting methodology that begins from the hold-up assumption embeds an unproven premise at the foundational step of the analysis.

FRAND exists not to discipline a patent holder but to preserve the innovation incentive of contributors who took real technical risk under conditions of genuine uncertainty, with no guarantee their approach would be selected. The bilateral framework of mutual obligations it creates is the mechanism through which that incentive is honored, not a ceiling on what a licensor can recover. Bowman Heiden and Justus Baron, writing in the Harvard Journal of Law and Technology, Vol. 38, No. 3 (2024), document a royalty gap of $7 to $28 billion annually as of 2021 in which SEP holders are systematically undercompensated relative to FRAND-compliant rates. (link) The baseline those figures use, ex ante contribution value rather than hold-up value, is the same baseline the Disney and Netlist Statements of Interest identify as the legally correct reference point under the rule of reason. The gap figure is contested at the margins, as all empirical measurements in active litigation fields are, but it is probative of a directional question: whether Rudin’s foundational framing creates a systematic methodological bias that runs against licensors, not whether any particular asserted rate is FRAND-compliant. On that directional question, a documented multi-billion-dollar annual gap is meaningful evidence that the framing is not analytically neutral.

In a genuine SDO, the sequence Rudin describes does not exist. There is no pre-existing patented technology waiting to be contributed. Engineers from Qualcomm, Nokia, Ericsson, Huawei, InterDigital, and dozens of other companies enter working groups together and develop technical solutions to defined problems collaboratively. They propose competing approaches. They debate tradeoffs. They vote, document objections, and arrive at consensus under conditions of genuine technical uncertainty. Nobody in the room knows at the outset which approach will be selected. The patents that ultimately read on the resulting standard are frequently filed during or after that process, on inventions conceived and refined inside the standards work itself, though patent applications on related technology filed prior to adoption are not uncommon. The technology and the standard come into existence through the same process. They are not separable artifacts.

The scale of that collaborative process in 3GPP is documented. For the period 2005 to 2014, approximately 350 firms made close to 400,000 unique contributions to 3GPP meetings. Roughly 30 percent of those contributions were accepted.² Implementer-aligned commentators have argued that large incumbent contributors arrive at working groups with pre-developed technology platforms and patent prosecution strategies coordinated to the standards calendar, making in-session invention a framing choice rather than a genuine description. That argument proves too much. A 30 percent acceptance rate across nearly 400,000 contributions from 350 firms, including proposals from the largest contributors in direct competition with each other, is not consistent with a process designed to ratify predetermined outcomes. Qualcomm, Nokia, and Ericsson are not aligned. They compete in the same working groups for the same slots in the same specifications. The record of contested proposals, documented objections, and non-selected technical approaches is precisely what a genuine competitive technical process produces. The continuation-portfolio timing practices that implementer advocates cite are a real and separate concern about how patents are prosecuted after a standard is adopted, and that question warrants its own analysis. It does not address whether the technical outcome of the working group process was determined before the meeting, which is the specific proposition the contribution data is cited to rebut.

That reality creates a serious problem for the rate-setting methodology Rudin describes. He defines “reasonable” as the rate a patent would command in a hypothetical negotiation where competitive alternatives exist, stripping out what he calls the “monopoly premium created by standardization.” That methodology requires that the patent’s value be analytically separable from the standard. Where technology and standard emerge from the same collaborative process, that separation is not available. For technology developed inside the working group, what exists at the moment of adoption is not a standalone inventin with a separable market value. What exists is a technical approach that competed against other approaches in the same process, any of which could have been selected instead. The appropriate ex ante question is what rate would have been agreed upon before adoption, when the alternative approaches were still viable options in the process. Rudin’s “monopoly premium” formulation assumes a premium exists to be stripped out. Where the contribution and the standard are the product of the same collaborative technical work, that assumption has no foundation.

The practical consequence of Rudin’s framing is that it imports the hold-up assumption as first principle without naming it as such. The two framings lead to different rate-setting methodologies, different default assumptions about who bears the burden of proving reasonableness, and different answers to what “fair” means in Fair, Reasonable and Non-Discriminatory.

Not the Same Destination

Chapter 5 is titled “Open Source and Standards: The Convergence.” Rudin’s argument is that over the past two decades the two worlds have converged “to the point where the line between them is often invisible,” with the same organizations hosting both, the same participants, the same governance challenges, and the same IP questions. Both are, in his framing, “fundamentally, structured collaboration among parties who don’t fully trust each other and don’t share interests in any deep way,” held together by the same creative tension between self-interest and collective benefit.

That framing collapses distinctions that matter enormously when a real dispute arises.

Rudin correctly identifies the first layer: an open source license covers contributed code, while a standards patent commitment covers the entire specification regardless of who contributed what. An independent implementer who reads open source code and builds their own version gets no patent coverage from the license. That is accurate as far as it goes.

It does not go far enough.

Governance structure. Formal standards development requires openness, balance across participant categories, lack of dominance by any single interest, consensus-based decision-making, documented resolution of objections, and a formal right of appeal. Those criteria define what separates genuine collaborative standard-setting from governance captured by a subset of participants. It is worth noting that even accrediting bodies have applied these criteria unevenly. ANSI’s position has historically been that governance — hence IPR policies that “govern” participation — do not warrant the same level of scrutiny as technical development processes, an approach that allowed the IEEE’s 2015 PatCom patent policy change to proceed under accreditation conditions that would not have survived rigorous application of the balance and consensus requirements to the policy process itself. That is a separate argument for a separate piece. The point here is narrower: whatever the floor of governance legitimacy requires in a formal standards body, a BDFL-governed open source project does not approach it. A significant proportion of major open source projects are governed by a single maintainer or a small committee operating on contribution-based authority with no due process obligations, no balance requirement, and no formal appeals mechanism. Linux, for decades governed by Linus Torvalds, is the canonical example. A BDFL-governed project meets none of the criteria that even minimally legitimate standards governance requires. The two governance structures are not alternative instruments for the same function.

Development process. In an SDO, a technical proposal enters a working group, competes against other proposals from parties with opposing economic interests, and survives or fails based on documented technical merit assessed through a consensus process with appeal rights. In an open source project, a pull request is accepted or rejected at maintainer discretion. The criteria, the accountability, and the institutional record are structurally different.

Patent coverage. The OSI’s OSD has ten criteria. They are licensing requirements, not governance or development standards. OSD criterion 7 states that the rights attached to the program must apply to all to whom the program is redistributed without the need for execution of an additional license by those parties. That criterion has been interpreted in OSI license-review discussions as potentially bearing on whether open source licenses must include a patent grant. The OSD is silent on patents, a product of its 1990s copyright-law origins. The patent coverage question under OSD criterion 7 remains genuinely unresolved, a point OSI itself has acknowledged and actively contested when FRAND-dependent companies have argued that open source licenses can be “copyright only.”¹ An SDO patent policy and an open source license are not covering the same ground through different mechanisms. In significant respects, the open source license covers no ground at all on the patent question.

The AOM/AV1 case. A practitioner trained on the convergence thesis will be least equipped to recognize a real-world problem when they encounter one.

AOM was not formed as a neutral technical governance vehicle for collaborative development. It was formed in 2015 by a founding membership that included Google, Mozilla, Microsoft (employer of David Rudin), Amazon, Netflix, Intel, and Cisco, large-scale downstream implementers who shared a direct financial interest in eliminating per-unit royalty costs for video codec technology. The AOM Patent License 1.0 grants each implementer a royalty-free license to each licensor’s necessary claims, on condition that the implementer grants its own necessary claims on the same royalty-free terms. The grant-back obligation is the mechanism. Any contributor who holds FRAND-encumbered patents and depends on licensing revenue faces a binary choice: accept royalty-free terms for all necessary claims or do not participate.

The predictable result of that structure is that FRAND-dependent innovators do not participate under the AOM framework. They retain their licensing positions and operate through independent patent pools and bilateral agreements outside the consortium entirely. The existence of AV1 patent pools operating outside the AOM RF license confirms the structural point: there are essential patents in AV1 that the RF license does not cover. The “open” label attached to AV1 is real at the code layer. At the patent layer, the picture is materially more complicated, and practitioners who assume otherwise on the strength of AOM’s governance narrative will advise their clients incorrectly.

Dina Kallay, serving as Deputy Assistant Attorney General for International, Policy and Appellate at the DOJ Antitrust Division, addressed exactly this dynamic in her September 2025 Concurrences keynote, titled “’That’s What F/RANDs Are For’ and Antitrust Implications When They’re Gone.”³ Her concern was specific, not general. She identified the breakdown of the FRAND-assured standards development ecosystem as an enforcement priority, with particular attention to two scenarios: collaborative standards that incorporate technologies carrying negative or missing FRAND declarations, and proprietary consortia whose policies are not FRAND-based. On the first scenario, her analysis tracks the AOM structure precisely. When a standard that is supposed to be open incorporates technologies whose patent holders have declined FRAND terms, or where the governance structure of the consortium makes FRAND participation economically irrational, implementers deploy the standard without any assurance that essential patent access will be available on competitive terms. That is not a generic competition concern about open labels. It is a targeted description of what AOM’s grant-back obligation produces for any contributor who depends on FRAND licensing revenue. The convergence thesis Rudin advances in Chapter 5 has no framework for that analysis. A practitioner working from his book will see AOM as an innovative governance model. A practitioner who has read Kallay’s keynote will ask a different set of questions about who designed the structure, who benefits from it, and whose ability to participate on commercial terms it eliminates.

In Chapter 17, Rudin observes that patent litigation is concentrated in the RAND standards world and nearly absent from open source, and acknowledges that “the reality is almost exactly backwards” from what his framework would predict. He does not resolve why. One reading of that absence is that royalty-free cross-licensing among AOM members resolved patent exposure efficiently, which is what the AOM structure was designed to accomplish. The more consistent reading, given the structural evidence, is different. The AOM Patent License 1.0’s grant-back obligation makes participation economically irrational for any contributor whose business model depends on FRAND licensing revenue. The predicted behavior under that structure is non-participation, not voluntary acceptance of RF terms. And the existence of AV1 patent pools operating entirely outside the AOM framework confirms that essential patents remain unlicensed under the RF license, which is precisely the outcome Deputy Assistant Attorney General Kallay identified as an antitrust enforcement concern: implementers deploying a standard without assurance that essential patent access is available on competitive terms. The absence of litigation is not evidence of a clean IP landscape. It is evidence of a participation structure that made FRAND-dependent contribution uneconomic from the outset. When there are no licensors in the room, there is no one to litigate against.

The Gap After Adoption

The three governance frameworks Rudin analyzes in Chapter 4 converge on process legitimacy. They govern the development of standards. They are silent on what happens after a standard achieves widespread commercial adoption.

Once a standard is broadly deployed and alternatives have receded from the market, the governance question shifts. The ANSI Essential Requirements no longer apply. The WTO TBT Decision has nothing further to say. EU Regulation 1025/2012 is silent. What governs from that point is the FRAND commitment made inside the standards body, and the bilateral framework of mutual obligations that commitment creates between the patent holder and every implementer who deploys the standard, including implementers who were never parties to the original standards process and who may have entered the market years or decades after the relevant technical work was complete.

That bilateral framework is not administered by the SDO. ETSI does not set rates. ETSI does not broker negotiations. ETSI does not resolve disputes. What ETSI’s IPR Policy does, and what Rudin’s warehouse metaphor cannot account for, is create a present-tense legal obligation at the moment of declaration. ETSI Clause 6.1’s “prepared to grant” language is not a procedural checkbox. It is a legal act with bilateral consequences: the patent holder commits to offer licenses on FRAND terms, and by accepting that commitment as the condition of standardization, the ecosystem creates a corresponding obligation on implementers to negotiate in good faith rather than to use the standards process itself as a mechanism for compressing or eliminating royalty obligations.

The Optis Wireless Technology LLC v. Apple Inc. proceedings now before the UK Supreme Court (UKSC/2025/0144 and UKSC/2025/0145), with a hearing scheduled for late June 2026, sit precisely in this space. The questions before the court go to the nature and enforceability of the obligations that arise from the FRAND commitment, who bears the burden of compliance, and what remedies attach when those obligations are not honored. Those are post-adoption questions. They are not answered by ANSI, TBT, or EU Reg 1025/2012. They are not addressed in Rudin’s Chapter 4. They are where the litigation lives, and they are what a practitioner trained solely on governance frameworks will be unprepared to analyze.

The Ecosystem Problem

The concern with a book like Coffee & Standards is not the book itself.

It is that well-written practitioner guides with embedded framing assumptions become baseline training. They travel into regulatory submissions, judicial briefing, and licensing negotiations. They shape which questions get asked and which frameworks get applied before a single argument is made on the merits.

The collaborative development error in Chapter 7 is not a footnote. It is the foundation of the FRAND rate-setting framework, and it imports the hold-up assumption as a first principle. The convergence thesis in Chapter 5 treats structurally distinct institutions as interchangeable, erasing differences in governance, development process, and patent coverage that define how disputes actually resolve. The post-adoption gap in Chapter 4 is the space where FRAND actually operates, and it is the space this kind of guide structurally cannot reach.

Sophisticated practitioners will read past these defaults. The audience most at risk is the one that will not, because no one told them to look.

Jim Harlan is an independent intellectual property practitioner and the author of Standards at Risk (standardsatrisk.com).

Footnote

1. Richard Fontana, “Is it time to revise the Open Source Definition?”, Opensource.com, September 30, 2020. Fontana notes that companies with FRAND-based SEP licensing models have advanced the view that open source licenses can be “copyright only,” and that the OSI has strongly opposed that position. The OSD’s silence on patents leaves OSD criterion 7 as the primary but contested textual hook for any patent coverage requirement. (link)

2. Kirti Gupta, “The Role of SMEs and Startups in Standard Development,” SSRN, July 2017. The contribution and acceptance data cited here are drawn from Gupta’s analysis of 3GPP participation for the period 2005 to 2014. (link)

3. Dina Kallay, “’That’s What F/RANDs Are For’ and Antitrust Implications When They’re Gone,” Remarks as Prepared for Delivery, Concurrences Dinner, New York, September 19, 2025. (link)

That dissent identified the real problem. The withdrawal statement says explicitly that it “does not prescribe how the Agencies will analyze antitrust issues raised by collaborations going forward.” The 1995 Antitrust Guidelines for the Licensing of Intellectual Property, on which the 2000 guidelines relied, had already been withdrawn. Businesses engaged in voluntary standards development and standards-essential patent licensing now operate without any surviving framework document.

I filed a comment on the public docket yesterday (Docket No. ATR-2026-0001). What follows is the argument I made, and why it matters.

What the inquiry is actually about

The DOJ/FTC joint inquiry asks three questions: what topics would benefit from additional guidance, what new technologies warrant clarity, and what significant legal and economic developments should inform revised guidelines. All three converge on the same domain: voluntary standards development through bodies like the Third Generation Partnership Project (”3GPP”), the Institute of Electrical and Electronics Engineers (”IEEE”), and the European Telecommunications Standards Institute (”ETSI”), and the Fair, Reasonable and Non-Discriminatory (”FRAND”) licensing framework that makes standards-based innovation commercially viable.

The agencies have an opportunity here that the 3-2 vote itself signals will not come around often. Two sitting FTC commissioners called for replacement guidance. The new administration has declared 6G a national security priority. The empirical record on two decades of theoretical harm claims has accumulated. This is the moment to get the framework right.

The harm theory that never found its evidence

The argument for aggressive antitrust oversight of FRAND licensing has circulated since the late 1990s. Its core claim is that SEP holders extract excessive royalties by threatening to block standard-compliant products, a dynamic called patent hold-up. Regulators have returned to this theory in jurisdiction after jurisdiction, generating enforcement actions, proposed regulations, and policy statements premised on it.

The empirical record has not confirmed it.

By early 2015, more than two dozen economists and lawyers had disputed the assumptions underlying the hold-up and royalty stacking conjectures. Anne Layne-Farrar’s 2014 survey of 15 years of economic literature for the OECD concluded that the studies had not shown hold-up to be a common problem. More recently, Bowman Heiden and Justus Baron, writing in the Harvard Journal of Law and Technology in 2024, examined the actual royalty picture and found not excessive extraction but a royalty gap: SEP holders are undercompensated for their patents by an estimated $7 to $28 billion annually as of 2021. They attribute the gap in significant part to implementer holdout, the strategic decision by firms to infringe rather than license, counting on the costs and delays of litigation to reduce or eliminate the payment they eventually make.

That finding inverts the standard regulatory narrative. It does not mean hold-up never occurs. It means guidance premised on hold-up as the dominant concern, while ignoring holdout, will systematically protect the wrong party.

What the record shows when guidance tilts toward implementers

The most direct evidence available is the IEEE’s own experience following the DOJ’s February 2015 BRL endorsing revisions to the IEEE’s intellectual property rights policy. New positive letters of assurance for 802.11 standards dropped 90 percent. Negative letters of assurance reached an all-time high in 2016. During 2015 through 2018, negative letters outnumbered new positive submissions. ISO/IEC formally raised concerns about IEEE standards entering the international fast-track process carrying unresolved negative assurances. ANSI declined to approve certain IEEE 802.11 standards as American National Standards.

In September 2022, IEEE rescinded the core provisions of the 2015 policy. The DOJ acknowledged in a subsequent filing that the policy may have discouraged participation, increased negative assurances, and contributed to the ANSI approval failures. That is not a theoretical prediction of harm. It is a documented outcome with named institutional consequences.

The agencies are not being asked to weigh in on a neutral policy domain. They intervened in this ecosystem before, watched the consequences accumulate, and are now soliciting input on what updated guidance should say. The prior intervention belongs in that analysis.

Why this is a 6G question, not just a licensing question

A February 2026 Boston Consulting Group analysis puts the economic stakes in concrete terms. 5G-enabled applications have already generated over $1 trillion in global economic impact. BCG projects cumulative value reaching $6.1 trillion by 2030 and $18.2 trillion by 2035. Those figures reflect the output of the same voluntary standards collaboration framework this inquiry addresses.

The same report documents that companies focused on core wireless technologies invest approximately 20 percent of revenues in R&D, comparable to pharmaceuticals and well above the rates committed by Apple, Amazon, and Google. BCG states directly: when IP enforcement becomes uncertain, it deprives these firms of the revenue that fuels R&D investment, undermining their ability to innovate and to compete for standards leadership. Regions that underinvest in R&D may find they have limited influence over standard development and must depend on foreign technology baselines.

The competitive context is not abstract. Western companies drove the development and standardization of technologies powering 2G through 5G. China has expanded its standards development role during the same period. South Korea and India have articulated explicit national targets for 6G standard-related patent shares. President Trump stated in December 2025 that 6G will be foundational to U.S. national security and pivotal to the development of artificial intelligence. Senate Commerce Committee Chairman Ted Cruz has warned that failure to lead in 6G would allow Huawei to create the backbone, “handicapping our efforts in other adjacent technologies like AI, quantum, and semiconductors.”

None of that is consistent with guidance that treats FRAND licensing as a presumptive antitrust concern.

What updated guidance should say

The comment I filed makes four specific asks.

Guidance should state that good-faith participation in open, voluntary standards development organizations is presumptively pro-competitive. The 2007 DOJ/FTC joint report on antitrust and intellectual property said as much. A new framework document should say it plainly and place the burden on those who claim otherwise to provide evidence, not theory.

Guidance should recognize FRAND licensing and joint licensing through patent pools as part of the incentive structure that makes standards-based innovation possible, not as categories of conduct presumptively suspect under antitrust law.

Guidance should address implementer holdout as a competitive harm on equal footing with patent hold-up. The Heiden and Baron royalty gap finding, the Gupta and Petrovčič systemic holdout documentation in the Berkeley Technology Law Journal, and the IEEE sequence described above all point in the same direction. An asymmetric framework that scrutinizes SEP holders while ignoring strategic non-licensing conduct by implementers will produce asymmetric outcomes. The comment also addresses a related tactic: some implementers have used the genuine complexity of value-chain licensing questions as a pretext for refusing to take any license at all, treating unresolved policy questions as a reason to avoid engagement rather than resolve them. Guidance that recognizes this dynamic will produce a more accurate picture of how FRAND licensing actually functions in practice.

Finally, guidance should be grounded in the empirical record as it now exists, not as it was theorized in the late 1990s. The withdrawal statement cited outdated analytical methods as a reason the 2000 guidelines no longer served their purpose. That logic applies with equal force to the underlying assumptions those guidelines carried forward from an earlier era of IP and antitrust thinking.

The dissenting commissioners asked for clarity. The filed comment argues for what that clarity should say.

The comment will be on the public docket at Docket No. ATR-2026-0001, but it is not as of the date of this article. (link, https://www.regulations.gov/docket/ATR-2026-0001)

Jim Harlan is an independent intellectual property practitioner with approximately 25 years of experience specializing in standards-essential patents, FRAND licensing, and standards policy. He publishes Standards at Risk at standardsatrisk.com.

What happened May 12, 2026: The UK Court of Appeal issued its judgment in Acer and Asus v Nokia Technologies OY, [2026] EWCA Civ 564 (link). Nokia’s arbitration offer was held to constitute a valid FRAND offer under the ITU Common Patent Policy. The Acer and Asus proceedings were permanently stayed. The planned FRAND trial will not proceed. Lord Justice Arnold, who authored the lead judgment, was on the panel.

Why it matters for InterDigital v Amazon: InterDigital is represented by Bird and Bird LLP in the Amazon dispute. The same Bird and Bird partner, Richard Vary, represents Nokia. The arbitration playbook is now legally validated and immediately available.

This dispute is not what it looks like. Amazon already holds a device-level license from InterDigital. InterDigital’s own Q4 2024 investor materials confirm “Amazon licensed for WLAN” in the CE and IoT channel. The ITC complaint targeting FireTV, Kindle, and Echo Show is the enforcement lever -- the mechanism to bring Amazon to the streaming licensing table. The eventual resolution will be a streaming services license under InterDigital’s Video Services licensing program, not a device license. This is InterDigital’s opening move in a $485 billion addressable market that is currently at zero ARR in its 2030 roadmap.

The prediction: InterDigital and Amazon announce binding ICC arbitration to determine the final terms of a streaming services license in Q3 or Q4 2026. The announcement itself is the resolution event -- exactly as the Samsung January 2023 and Lenovo Q4 2024 announcements were. InterDigital begins conservative revenue recognition immediately. All active litigation is stayed by consent. The arbitral tribunal determines the final streaming services rate 18 to 24 months later on a confidential basis. No public FRAND rate enters the record.

The confidence: 60 percent in Q3-Q4 2026 for the arbitration announcement. This is the correct endpoint -- not Amazon accepting a final streaming license, but Amazon agreeing to have a neutral ICC tribunal determine whether and on what terms streaming services require a separate patent license.

The key driver: The May 28, 2026 UPC Court of Appeal hearing on the AILI penalty orders. The critical question is whether any modified AILI retains enough coercive force -- in penalty quantum and geographic scope -- to override Amazon’s internal resistance. A modified AILI that preserves the core 50 million euro penalty structure tips toward settlement. A modified AILI that reduces the penalty cap below approximately 20 million euros or narrows geographic scope to exclude major EU markets likely tips toward the UK trial proceeding.

The threshold question that has not been connected to the Nokia EWCA arbitration analysis: Before the Nokia EWCA framework applies, someone must determine whether any asserted patents are ITU-T declared SEPs. Based on patent-level analysis, the identified portfolio reads primarily on AV1 and HDR -- not HEVC decoder-side SEPs. AV1 was developed by the Alliance for Open Media, not an SDO. The ITU Common Patent Policy was never declared on AV1 patents. And even for patents that might qualify, “would be required to implement” under the ITU policy is governed by Swiss law and undefined -- to the best of my knowledge courts have not determined whether it means technically required or commercially required. The encoding/decoding essentiality question is formally in dispute before Judge Meade, with the WilmerHale FRAND Quarterly noting in March 2026 that the Amazon proceedings center on “whether the ITU-T RAND obligation applies to both encoding and decoding patents or only the latter.” (link)

The methodology: Part 5 ran six separate swarm iterations. Confidence across the six runs: 63, 54, 61, 64, 47, and 60 percent. The volatility is itself an analytical signal -- the confidence number is sensitive to how you frame the endpoint. The definitive run uses the correct two-layer framework: Layer 1 is the arbitration announcement (the prediction endpoint), Layer 2 is the final arbitral award 18 to 24 months later.

What the Nokia EWCA Decision Actually Holds

The judgment in Acer v Nokia is more nuanced than the headline framing suggests.

Nokia owns a portfolio of SEPs declared essential to ITU-T H.264/AVC and H.265/HEVC video decoding standards. Acer and Asus filed in the UK Patents Court seeking a court-determined global FRAND rate. Nokia challenged jurisdiction and offered arbitration instead. Justice Mellor at first instance rejected Nokia’s position and granted interim license declarations. Nokia appealed.

The Court of Appeal reversed. Three holdings matter for the Amazon dispute.

1. Nokia’s arbitration offer constitutes a valid FRAND offer capable of acceptance. An implementer who refuses a genuine arbitration offer and insists on court-determined terms is not a willing licensee in the relevant sense.

2. The ITU Common Patent Policy does not require a SEP holder to submit to court jurisdiction. The FRAND commitment is a commitment to license on reasonable terms -- not a commitment to have those terms set by a national court.

3. The stay is conditioned on Nokia agreeing that statements of case, disclosed documents, and evidence from the English proceedings stand as arbitration materials, and that costs incurred are costs in the arbitration.

The Wolters Kluwer Patent Blog (link) correctly identified the deeper doctrinal point the day after the judgment: the Court of Appeal did not compel arbitration. It recognized that Nokia’s offer embedded arbitration within the performance of a contractual undertaking. The FRAND obligation is a contractual obligation to make access available on objective terms. Arbitration does not displace it -- it may concretize it.

Nokia also offered an immediate interim license alongside the arbitration offer. The Court expressly left open whether a mere offer to arbitrate, without an accompanying interim license, would suffice. InterDigital therefore needs to include an immediate interim license in its offer to Amazon to maintain the full protection of the Nokia precedent.

What Arnold LJ expressly left open:

Whether the Adjustable License’s litigation abandonment term -- requiring parties to stay or abandon their own litigation as a condition of the license -- could be used to distinguish a future offer was left open at paragraphs 28 and 92. Whether an implementer with a legitimate and substantiated objection to the proposed arbitral mechanism could resist the stay was also left open. Amazon holds patents and has active UK and Brazilian proceedings. These are its primary avenues of resistance.

The Threshold Question: Are These Patents Even SEPs?

Before the Nokia EWCA arbitration framework applies to the Amazon dispute, a threshold question must be resolved that has not yet been connected to this decision in public commentary.

The AV1 FRAND gap has been noted. IP Fray observed in February 2026 (link) that “InterDigital definitely didn’t participate in VP9 and AV1 standard-setting, so there are no FRAND pledges in place with respect to those standards.” The WilmerHale FRAND Quarterly noted in March 2026 that the Amazon UK proceedings center on “whether the ITU-T RAND obligation applies to both encoding and decoding patents or only the latter” -- confirming the essentiality question is formally in dispute before Judge Meade.

What has not been connected to the Nokia EWCA decision is the consequence: if the asserted patents are not ITU-T declared SEPs, the procedural trap created by Acer v Nokia may not close on Amazon at all. The unwillingness finding flows from refusing a valid FRAND offer. If the patents are not FRAND-encumbered, there is no valid FRAND offer to refuse.

Based on patent-level analysis of the identified US proceedings:

The ITC and Delaware companion patents (US 10,741,211, US 9,747,674, US 8,363,724, US 8,681,855, and US 11,917,146) read primarily on AV1 and HDR, not on HEVC decoder-side implementations. AV1 was developed by the Alliance for Open Media -- not by ITU-T. No ITU Common Patent Policy declaration was made on AV1 patents. InterDigital joined the SISVEL AV1/VP9 pool in March 2020 but is no longer a member, having chosen direct bilateral assertion at rates unconstrained by the SISVEL pool ceiling of approximately 0.11 euros per device.

The Eastern District of Virginia patents (US 8,149,338, US 11,252,435, US 12,149,734, and US 12,143,606) similarly read on AV1 and HDR. The Western District of Texas patents are definitively non-SEP Edgio-acquired content delivery patents -- US 7,921,259, US 10,116,565, US 8,745,128, US 9,015,416, US 8,868,701, and US 8,583,769 -- with no connection to any video codec standard.

US 8,363,724 has additional corroboration: formerly owned by Technicolor, who self-declared it essential to H.265, and previously in the SISVEL pool. InterDigital’s ITC complaint in Investigation 337-TA-3869 alleges each of the five ITC patents is essential to ITU-T H.265 -- that allegation, made in a federal enforcement proceeding, is the primary evidentiary basis for the Category 1 characterization.

InterDigital’s SISVEL pool departure reflects its preference for direct bilateral assertion unconstrained by pool ceilings. Some pools reward active litigators through enhanced distribution mechanisms. The pool departure maximizes direct recovery potential while concentrating validity risk on individual patents. The Unified Patents IPR challenge to US 8,363,724 illustrates that exposure.

The ITU Common Patent Policy essentiality gap:

The ITU policy’s “would be required to implement” threshold is governed by Swiss law and is undefined. ETSI uses “necessary” -- interpreted as technically necessary at the claim level. The ITU policy leaves open whether “required” means technically required or commercially required. Swiss courts and arbitral tribunals have not resolved this, to the best of my knowledge. Non-US courts including UPC Local Divisions and any ICC arbitral tribunal will need to address it independently.

The pro-licensor consequence:

If the AV1 and HDR patents fall outside the FRAND framework, InterDigital can assert them on uncapped commercial terms with no FRAND ceiling. FRAND is Amazon’s preferred battlefield because it potentially caps the rate. If the patents escape FRAND characterization, InterDigital seeks market-rate licensing unconstrained by any SDO declaration.

The Arbitration Weapon and Its Catch

InterDigital has a clear playbook. It can offer Amazon portfolio-level ICC arbitration, structured similarly to its Samsung and Lenovo agreements, and apply to the UK High Court for a stay of the FRAND proceedings. Under the May 12 EWCA decision, that application has a strong legal foundation -- provided the offer includes an immediate interim license and avoids the litigation abandonment term Arnold LJ flagged at paragraphs 28 and 92.

Amazon holds patents and has active affirmative proceedings in the UK and Brazil. If InterDigital replicates Nokia’s litigation abandonment term, Amazon has a distinguishing argument. A clean offer -- ICC arbitration, portfolio scope covering all patent categories, no litigation abandonment condition, immediate interim license, neutral seat and panel selection -- would be significantly harder to resist.

The Procedural Trap: Good Faith Refusal and the Unwilling Licensee Risk

The May 12 EWCA decision created a structural asymmetry that Amazon’s legal team will have identified immediately.

InterDigital makes Amazon a clean arbitration offer. Amazon believes in good faith the offer is not FRAND -- perhaps because the scope risks cascading into AWS and Prime Video infrastructure. Amazon refuses. The UK court later finds the offer was a valid FRAND offer. Amazon is retroactively characterized as an unwilling licensee -- not for refusing a license, but for refusing what the court determined was a FRAND offer. Amazon loses willing licensee status without ever getting a merits determination on the rate.

Amazon has three strategic responses.

1. Accept the arbitration offer and challenge rate, scope, and structural terms within the arbitral process. Unwilling licensee risk removed. Forum ceded to InterDigital. Critically, Amazon is NOT accepting that streaming services require a separate license -- it is agreeing to have a neutral tribunal determine that question. This is a materially lower bar than accepting a final license.

2. Preemptively file in the UK court for a determination that any InterDigital arbitration offer is not FRAND before the offer is formally made. The six-week procedural race.

3. Accept the principle of arbitration but challenge the specific structural terms of InterDigital’s offer. Arnold LJ left this open at paragraphs 28, 74, 87, and 92. The most surgical path.

The asymmetry did not exist before the May 12 judgment. It is a genuine shift in InterDigital’s favor regardless of which option Amazon pursues.

The Two-Layer Framework: Announcement as Resolution

The most important analytical refinement in Part 5 is the distinction between two resolution events: the announcement of binding arbitration and the final arbitral award. These occur years apart. For prediction purposes, the announcement is the event that matters.

Layer 1 -- The announcement (the prediction endpoint):

InterDigital and Amazon announce binding ICC arbitration to determine the final terms of a streaming services license. This single announcement stays all active litigation by consent, begins InterDigital conservative revenue recognition immediately, removes the September 2026 UK trial risk for both parties, and gives InterDigital proof of concept for its 2030 ARR roadmap. Amazon is not accepting that streaming services require a separate license -- it is agreeing to have a neutral tribunal determine that question. All of Amazon’s arguments on rate, scope, streaming licensing legitimacy, AWS pass-through liability, and cascade externality cost are preserved for the tribunal.

Samsung precedent: January 1, 2023 announcement. Q1 2023 conservative revenue recognition began. The Q1 2023 10-Q stated: “We believe that it is likely the arbitration award will exceed the conservative estimate and require a true-up at that time.” July 2025: $1.05 billion ICC award, 2.5 years after announcement.

Lenovo precedent: Q4 2024 announcement. Revenue recognition began immediately. Award expected year-end 2026.

Layer 2 -- The arbitral award (18-24 months later):

The ICC tribunal determines the final terms of the streaming services license. Confidential. No public precedent. The tribunal addresses the streaming licensing legitimacy question, the rate methodology for SVOD and AVOD, and the AWS pass-through liability scope -- all on a confidential basis that protects both parties from public precedent.

Why Layer 1 has higher probability than prior runs captured:

Amazon is agreeing to arbitrate -- not to accept InterDigital’s streaming licensing theory. The bar is materially lower than accepting a final license. The Nokia EWCA procedural trap means refusing a properly structured arbitration offer risks the unwilling licensee finding. InterDigital’s institutional investor pressure and ARR target create board-level urgency to announce an arbitration agreement before the Q3 2026 earnings call. And crucially, the AWS pass-through liability uncertainty actually favors Layer 1 -- if Amazon’s legal team concludes AWS pass-through liability is material, announcing arbitration to contain the scope is better than litigating through a public UK judgment that could establish broader liability. Arbitration scope is negotiated and confidential. Judicial scope is public and precedential.

InterDigital’s Strategic Imperative: Streaming

InterDigital’s September 2024 investor day announced a target of $1 billion in annual recurring revenue by 2030 across three channels. Channel 1 -- smartphone -- currently at $491 million ARR, target $500 million. Channel 2 -- consumer electronics, IoT, and automotive -- currently at $97 million ARR, target $200 million. Channel 3 -- streaming and cloud services -- currently at zero ARR, target $300 million or more.

Amazon is the first major enforcement action in Channel 3. No major streaming platform has been licensed under the Video Services program yet. InterDigital’s Q4 2025 earnings call stated: “Litigation against major streaming services is progressing favorably with positive injunctions achieved.”

A resolution that produces no streaming services license at all would signal to every other SVOD and AVOD platform that Prime Video is not licensable. That signal would undermine the entire Channel 3 thesis. InterDigital cannot accept that outcome. Amazon knows this. The negotiation is not about whether there will be a streaming component -- it is about the rate structure, scope, and whether the license is structured to minimize its value as a public reference point for Netflix, Disney Plus, and every other platform.

The Disney precedent is directly relevant. InterDigital filed against Disney as a streaming service in the same Video Services program enforcement campaign. Disney streaming services were injuncted in Brazil and Germany. Munich imposed 550,000 euros in fines on Disney in January 2026 for injunction non-compliance. Disney is the direct comparable for Amazon’s litigation posture and settlement calculus. If Disney announces arbitration before Amazon does, Amazon becomes the last holdout against an established licensing program rather than the industry’s test case defender.

The Three-Category Patent Framework

The portfolio InterDigital is asserting contains three distinct patent categories.

Category 1: Plausibly FRAND-committed SEPs (approximately 5 of 15 identified US patents).

The ITC/Delaware patents. InterDigital’s ITC complaint in Investigation 337-TA-3869 alleges each is essential to ITU-T H.265. US 8,363,724 has additional corroboration from Technicolor’s prior H.265 self-declaration. Amazon cannot cleanly refuse arbitration on the grounds the portfolio is “not FRAND” because these five patents almost certainly carry FRAND obligations.

Category 2: Contested implementation patents with uncertain FRAND status (approximately 4 of 15).

The Eastern District of Virginia patents -- AV1 and HDR. InterDigital’s position is that encoder-side patents are not FRAND-committed because HEVC specifies only decoder behavior. FRAND status legally unresolved.

Category 3: Definitively non-SEP content delivery patents (approximately 6 of 15).

The Western District of Texas Edgio-acquired patents. No connection to any video codec standard. No FRAND obligation. Targets AWS and Prime Video infrastructure. Uncapped damages potential.

The critical clarification: Amazon already holds a device-level license from InterDigital covering WLAN and related device technologies. InterDigital’s Q4 2024 investor materials confirm “Amazon licensed for WLAN.” The ITC complaint targeting FireTV, Kindle, and Echo Show is the enforcement lever -- not the subject of the eventual license. The resolution will be a streaming services license under the Video Services program.

The bundled arbitration path covering all three categories in a single ICC proceeding is more efficient for Amazon than fighting each category separately across three different legal frameworks simultaneously.

The German Enforcement Mechanism and UPC Compliance Failures

The German Regional Court proceedings are a primary business-pressure mechanism and the UPC Mannheim compliance failures are a distinct and compounding enforcement risk.

InterDigital filed three separate infringement proceedings in Munich and Mannheim Regional Courts against Amazon in November and December 2025. Trials are expected in Q3 and Q4 2026.

The UPC Mannheim compliance record is aggressive. February 6, 2026: Mannheim court requested explanations regarding the UK hearing. February 11, 2026: order stating that claims for damages in the UK proceedings could improperly interfere with UPC jurisdiction -- constraining the settlement structure itself. February 27, 2026: Mannheim found Amazon in breach of the ASI and required a formal declaration before the UK High Court. March 9, 2026: Judge Tochtermann questioned whether Amazon’s UK declaration had binding legal effect. Both the German ASI and the UPC AILI were issued ex parte -- Amazon could not pre-empt them with UK applications, increasing exposure to compounding enforcement actions.

Amazon has already demonstrated a ceiling on its litigation tolerance by withdrawing its UK damages claim under UPC coercive pressure. That is not a trivial concession and signals that Amazon’s attrition posture has limits. The sequential enforcement escalation -- breach findings, compliance orders, financial penalties -- creates a live contempt exposure that is not merely theoretical.

Munich imposed 550,000 euros in fines on Disney in January 2026 for injunction non-compliance. That enforcement template makes Amazon’s European exposure more quantifiable and less deferrable than prior analysis assumed.

Amazon’s Structural Resistance and the Streaming Industry Collective Action Dynamic

Amazon’s resistance to any streaming services license is structurally more durable than a simple cost-benefit calculation suggests -- and for a reason the prior analysis missed entirely.

If Amazon accepts a streaming services license, it does not just bear its own royalty burden. It sets an industry-wide MFN-anchored ceiling that harms every SVOD and AVOD competitor simultaneously. Netflix, Apple TV Plus, Disney Plus, and Paramount Plus all compete with Prime Video. Amazon accepting a streaming patent license baseline creates a cost structure that InterDigital can then use as a floor in negotiations with every other streaming platform. Amazon is effectively litigating on behalf of the entire streaming industry, and the industry has a material financial interest in Amazon holding the line.

This streaming industry collective action dynamic -- Netflix, Apple, and other SVOD platforms informally coordinating to support Amazon’s resistance -- is structurally invisible in the public litigation record. But it raises Amazon’s effective settlement floor above what InterDigital’s institutional shareholder pressure requires InterDigital to accept. If this coordination is real and Amazon internalizes the industry-wide precedent cost, the gap between the parties’ reservation prices may be unbridgeable before the September 2026 trial.

The AWS pass-through liability dimension compounds this. AWS CloudFront and Elemental Media Services customers use Amazon’s infrastructure to deliver their own streaming content. If a streaming services license attaches to video delivery infrastructure, the royalty obligation could extend beyond Prime Video to reach AWS customers -- a liability exposure orders of magnitude larger than Prime Video alone. Paradoxically this uncertainty could either accelerate settlement (Amazon wanting to contain scope through confidential arbitration) or harden resistance (any license risks activating the pass-through problem). The direction depends on Amazon’s internal quantification of that exposure.

The DG COMP flank:

InterDigital’s subsidiary DRNC Holdings acquired the six Edgio content delivery patents from Edgio Inc.’s bankruptcy in January 2025 -- seven months before Amazon filed in the UK. The pre-dispute timing, targeting Amazon’s cloud infrastructure specifically, is a potential basis for a complaint to DG COMP under Article 102 TFEU. The UPC Mannheim notified the European Commission on December 24, 2025. A DG COMP complaint does not produce immediate relief but gives Amazon additional leverage in negotiating any arbitration scope.

The Revised House View

Prediction: InterDigital and Amazon announce binding ICC arbitration to determine the final terms of a streaming services license in Q3 or Q4 2026. The announcement is the Layer 1 resolution event -- stays all litigation, begins conservative revenue recognition for InterDigital, removes the September 2026 UK trial risk for both parties. The Layer 2 arbitral award follows 18 to 24 months later on a confidential basis, covering Prime Video delivery as an SVOD and AVOD streaming platform at a rate deliberately structured to minimize its value as a public industry reference point. No public FRAND rate enters the record. The streaming industry collective action dynamic means Amazon accepts arbitration -- not because it concedes the streaming licensing theory -- but because confidential ICC arbitration scope is better than a public judicial determination that could activate the AWS pass-through liability problem.

Alternative Outcome: The September 2026 UK High Court FRAND trial proceeds to a full judgment, issuing a global rate determination in late 2026 or early 2027. This occurs if the UPC Court of Appeal on May 28 vacates or materially narrows the AILI, reducing the penalty cap below the level that makes Amazon’s cost-of-delay calculus favor settlement, and if the Dolby California DJ complaint produces an adverse claim construction on the five Amazon-asserted patents before September 2026, collapsing a material portion of InterDigital’s patent position. The judicially-set rate excludes encoder-side patents and becomes a structurally damaging precedent for InterDigital’s Video Services program. Probability: approximately 30 to 35 percent.

Key Driver: The May 28, 2026 UPC Court of Appeal hearing on the AILI penalty orders. A binary survival or vacatur is less likely than a partial modification. The critical question is whether any modified AILI retains enough coercive force -- in penalty quantum and geographic scope -- to override Amazon’s resistance. A modified AILI preserving the core 50 million euro penalty structure tips toward settlement. A modified AILI reducing the penalty cap below approximately 20 million euros or narrowing geographic scope to exclude major EU markets likely tips toward trial.

Residual Uncertainty: Whether the streaming industry collective action dynamic -- Netflix, Apple, and other SVOD platforms informally coordinating to support Amazon’s resistance -- is sufficiently organized and financially material to raise Amazon’s effective settlement floor above what InterDigital’s institutional shareholder pressure requires InterDigital to accept. This factor is structurally invisible in the public litigation record and cannot be resolved from available information.

Main Disagreement: Whether InterDigital will deploy the Nokia EWCA arbitration stay mechanism as a forcing function for settlement, or withhold it because the UK trial trajectory on encoder-side essentiality produces a judicial rate at or above InterDigital’s arbitration floor, making court-driven resolution preferable. The Adversarial Critic argues the UK trial is more likely to proceed to judgment than the consensus assumes because InterDigital’s incentive to stay the proceedings depends on a private rate calculation no external analyst can observe.

Confidence: 60 percent in Q3-Q4 2026 arbitration announcement.

What Changed Across Five Parts

The directional prediction has held across all five iterations: confidential global resolution, not a public judgment. The mechanism, timing, and confidence have evolved significantly.

Part 1 predicted Q4 2026, driven by the UPC penalty. Confidence: 52 percent.

Part 2 corrected the Delaware error. Confidence: 54 percent.

Part 3 added the validity challenge layer and ITU Common Patent Policy analysis. Confidence: 68 percent.

Part 4 anchored to the Q1 2026 10-Q, added the Samsung ICC award, Dolby California DJ action, and German court trial schedule. Confidence: 61 percent.

Part 5 ran six swarm iterations. The final definitive run incorporates the Nokia EWCA judgment, the Samsung and Lenovo two-layer announcement framework, the patent fraction stress test, InterDigital’s 2030 ARR streaming channel roadmap, the Amazon existing device license correction, and the streaming industry collective action dynamic. Confidence: 60 percent in Q3-Q4 2026 arbitration announcement.

The confidence volatility across Part 5 runs -- 63, 54, 61, 64, 47, 60 -- is itself an analytical signal. The number is sensitive to how you frame the endpoint. A final streaming license has a different probability than a binding arbitration announcement. The Samsung and Lenovo precedents establish that the announcement is the correct endpoint. Under that framing, 60 percent is the honest convergence.

The most important analytical development across the series is the identification of the streaming industry collective action dynamic and the two-layer announcement framework. This dispute is not a device patent enforcement action. It is InterDigital’s opening move in a $485 billion addressable market licensing channel that does not yet exist in its ARR base. The Samsung and Lenovo precedents show how InterDigital wins: not by extracting a final rate, but by announcing that the rate will be determined by a neutral tribunal -- and booking the revenue immediately.

A Note on the Methodology

Part 5 ran six separate swarm iterations before producing the final output published here. Each iteration added new analytical material. The six runs produced confidence levels of 63, 54, 61, 64, 47, and 60 percent. The volatility reflects genuine uncertainty about endpoint definition, not analytical instability. The final run uses the correct two-layer framework anchored to the Samsung and Lenovo arbitration announcement precedents.

The definitive Part 5 input file reached approximately 145,000 characters, the largest of the series. The swarm ran on Claude Sonnet 4.6 via the Anthropic API. The Anthropic API experienced a sustained Sonnet 4.6 capacity incident on May 12, 2026 during North American business hours. All six Part 5 runs completed successfully after the incident resolved.

The full Python implementation including all pre-processors and the ten-agent swarm architecture is available at github.com/theharlans/standards-at-risk.

Sources: [2026] EWCA Civ 564, Acer Incorporated and Asus v Nokia Technologies OY and others, judgment May 12, 2026, judiciary.uk. InterDigital Inc. Form 10-Q for the quarterly period ended March 31, 2026, filed April 30, 2026, SEC EDGAR. InterDigital FY2025 Annual Report, SEC EDGAR. InterDigital Q4 2024 supplemental materials, February 6, 2025, SEC EDGAR. InterDigital Q3 2025 supplemental materials, October 30, 2025, SEC EDGAR. InterDigital press release, “InterDigital and Samsung conclude arbitration and announce new license agreement,” July 29, 2025, ir.interdigital.com. InterDigital Form 10-Q for the quarterly period ended March 31, 2023, SEC EDGAR. Matthieu Dhenne, “Come Together? Acer v Nokia and the Contractual Turn of FRAND Arbitration,” Kluwer Patent Blog, May 13, 2026. WilmerHale FRAND Quarterly, March 2026. IP Fray, February 2026. UK Supreme Court case record UKSC/2025/0058, Tesla v InterDigital and Avanci, heard April 28 to 30, 2026, decision forthcoming. All analysis is the author’s own. This is not legal advice.

For those who follow standards, standard-essential patents (”SEPs”), and the licensing ecosystem that underpins global technology commerce, there is a quieter but consequential development worth examining. Section H of the Report, titled “Intellectual Property and Standards,” takes meaningful positions on SEP enforcement. Some of what it says aligns directly with arguments the pro-licensor, pro-innovator community has been making in formal submissions. Some of what it omits is equally instructive.

What I Submitted and Where the Report Aligns

Earlier this year I filed comments with USTR in the 2026 Special 301 Review process, focused on the intersection of SEP licensing, implementer holdout, and trade policy. Several of those arguments appear in the final Report, at least in directional form.

On Licensing Negotiation Groups (”LNGs”): my submission identified state-facilitated licensee negotiation groups as risks for coordinated buyer behavior, functioning as buyer cartels that impose collective rate pressure and erode the value of patent rights. Section H names LNGs explicitly and uses language invoking both holdout and sub-fair, reasonable and non-discriminatory (”FRAND”) pricing as specific concerns. That is direct alignment.

On rate-setting sovereignty: my submission argued that foreign courts setting global FRAND rates for portfolios that include U.S. patents constitutes interference with U.S. patent holders’ domestic enforcement rights. Section H frames this in nearly identical terms, grounding the concern in the traditional territoriality of patent rights. Also aligned.

On injunction restrictions: my submission argued that categorical limits on injunctive relief in the SEP context deprive innovators of credible enforcement and function as de facto price controls. The Report identifies this trend by name and frames it as a threat to American innovation leadership, economic competitiveness, and national security.

USTR does not cite individual submissions. That is standard practice. But the directional consistency between what the pro-licensor community submitted and what appears in Section H is not accidental. It reflects where sustained, substantive engagement with the policy process has been moving the conversation.

Where the Report Falls Short

Recognizing progress does not require overlooking the gaps. Three stand out.

First, the Report names LNG holdout but does not develop a general theory of implementer holdout as a standalone trade distortion. Holdout exists independently of organized negotiation groups. A single implementer using standardized technology while deferring licensing indefinitely, leveraging litigation cost asymmetries and regulatory uncertainty to depress royalties, is engaging in holdout just as surely as a formally organized LNG. Treating holdout as a feature of LNG behavior rather than a systemic practice leaves the stronger argument on the table.

Second, the Report identifies sovereign rate-setting and injunction restrictions as problems but offers no affirmative FRAND economics to support the critique. The word “FRAND” appears in the LNG passage but is never analyzed. What makes a rate FRAND? What is the relationship between expected licensing returns and decisions to invest in research and development and contribute technology to standards development organizations (”SDOs”)? What happens to those incentives when FRAND is interpreted in practice as the lowest rate an implementer will accept under litigation pressure? None of this is addressed. The concerns are named without any analytical architecture behind them.

Third, the Report’s silence on two significant recent European developments is a genuine omission. The European Union proposed a sweeping SEP Regulation that would have imposed mandatory essentiality checks, aggregate royalty benchmarks, and new procedural constraints on SEP licensing. That proposal was withdrawn in 2024 after sustained opposition. Separately, the European Commission published final Technology Transfer Guidelines (”TTGs”) in April 2026. An earlier draft of those guidelines had included a safe harbor provision for LNGs. The final version removed it. Both developments represent concrete pro-licensor outcomes in a jurisdiction the Report elsewhere treats with concern. Neither is acknowledged. A report genuinely engaged with global SEP dynamics would have noted that the regulatory overcorrection narrative has encountered real resistance in Brussels, and that the resistance has produced documented results.

What the Report Actually Says, and the Takeaway

Section H opens by framing efficient and predictable SEP licensing as critical for both innovators contributing technology to standards and companies needing licenses to implement them. It then states plainly that American innovation leadership, economic competitiveness, and national security are threatened by actions that undermine effective patent enforcement. That framing, at the top of the section, sets a pro-licensor tone that runs through everything that follows.

The specific trends identified are: overbroad ASIs that prevent U.S. entities from enforcing patent rights in other jurisdictions; foreign courts asserting authority to set worldwide FRAND rates without consent from both parties; categorical prohibitions on injunctive relief in the SEP context; and LNGs enabling technology purchasers to act jointly to lock in potentially sub-competitive or sub-FRAND prices, or to agree to engage in holdout.

On China specifically, the Report identifies the State Administration for Market Regulation (”SAMR”) Anti-Monopoly Guidelines in the Field of Standard Essential Patents, issued in final form in November 2024, as a vehicle for potential misuse of anti-monopoly enforcement to favor domestic companies. It notes that Chinese courts appear increasingly interested in asserting jurisdiction over SEP disputes globally, and that senior Chinese judicial and political authorities have described ASI issuance as serving the overall work of the Chinese state. That is pointed language for a trade document.

The honest assessment is this: the 2026 Special 301 Report is the best SEP-related language USTR has produced in its annual reporting. It is also, precisely because it gets closer than before, a clear illustration of how far U.S. trade policy remains from having a coherent standards strategy. The Report is an enforcement document. It identifies harmful practices and applies pressure. It is reactive by design.

What the ecosystem needs is an affirmative framework: one that treats voluntary, private-sector-led standards development as strategic infrastructure; that connects R&D investment, SEP licensing returns, and reinvestment in next-generation standards as a coherent innovation cycle; and that names implementer holdout as a trade distortion in its own right.

The 2026 Report signals that the United States is beginning to see the problem clearly. The policy architecture to address it does not yet exist. That gap is what Standards at Risk is here to document.

Standards at Risk covers SEP licensing, FRAND policy, and the intersection of innovation and trade. Subscribe for analysis that goes beyond the headlines.

What this is: Part 4 of a series using a ten-agent AI reasoning system to predict the outcome of the InterDigital v Amazon video codec and streaming patent dispute. This run uses InterDigital’s Q1 2026 10-Q filed April 30, 2026 as the primary seed document -- the first time an SEC-filed primary source has anchored the analysis.

The prediction: A bilateral global patent license agreement in Q3 or Q4 2026, before the UK High Court issues a formal binding RAND judgment. Confidence: 61 percent.

The key driver: The May 28, 2026 UPC Court of Appeal hearing on the Anti-Interim-License Injunction (“AILI”) scope. Whatever it decides accelerates resolution. An upheld AILI compels Amazon to settle to avoid cascading injunction exposure during the Q4 holiday cycle. A narrowed AILI removes InterDigital’s most aggressive enforcement tool and increases its own incentive to settle before losing further ground.

What the 10-Q changed:

The Samsung $1.05 billion ICC arbitration award is the most important new fact. It covers cellular SEP patents in the smartphone context, not video codec patents in the streaming device context, so it is not a directly applicable rate comparable for the Amazon proceedings. What it does signal is the scale of awards InterDigital is willing to pursue through arbitration and its demonstrated willingness to litigate rather than settle early. Samsung challenged the award in December 2025. If the challenge succeeds before September 2026, InterDigital’s internal settlement posture may soften regardless of the technology category difference.

Dolby escalated from a PTAB IPR to Article III federal court in March 2026, seeking declaratory judgment on five patents also being asserted against Amazon. That is a more immediate validity threat than any prior analysis captured.

German court proceedings against Amazon are scheduled for trial in Q3 and Q4 2026 -- independent of the UPC and running on their own timeline. Prior analysis missed this entirely.

The DOJ filed a Statement of Interest in the Disney Delaware antitrust case in October 2025 -- but contrary to the prior analysis framing, it sided with InterDigital rather than against it. The DOJ steered the dispute away from antitrust law, reducing rather than increasing the regulatory flanking risk to InterDigital.

The AV1 correction: The prior analysis modeled AV1 adoption as a risk to InterDigital’s portfolio. That was wrong. InterDigital holds AV1 patents and is actively asserting them against Amazon. Amazon devices capable of decoding AV1 or VP9 content are named as accused instrumentalities in the Delaware complaint. AV1 adoption expands InterDigital’s enforcement surface, it does not erode it.

Why confidence dropped from Part 3 (68%) to Part 4 (61%): A richer fact base surfaces previously invisible risks. The Samsung award challenge and the Dolby California proceedings introduce genuine uncertainty that was not in the Part 3 input. It also surfaces corrections -- the AV1 risk framing and the DOJ framing were both wrong in ways that the primary source data revealed. More information does not always increase confidence -- sometimes it reveals what you did not know you did not know.

A Different Kind of Seed Document

Parts 1 through 3 of this series were built from a single JUVE Patent article supplemented by live PACER docket pulls and PTAB API data. Each part improved on the prior one by correcting errors and adding data sources. But all three shared a foundational limitation: the primary seed document was journalism, not primary source material.

Part 4 is different. On April 30, 2026, InterDigital filed its Q1 2026 10-Q with the SEC. The litigation section runs to approximately 22,000 words covering every active proceeding across every jurisdiction against every defendant. It is authoritative, comprehensive, and verified under penalty of law. It covers Amazon, Disney, Dolby, Hisense, TCL, Lenovo, Samsung, Transsion, and Tesla simultaneously.

Running the ten-agent swarm on this document produced the richest output of the series. The confidence number moved to 61 percent. More importantly, three new analytical contributions emerged that were absent from any prior run, and one latent factor surfaced that has received no public attention in the commentary on this dispute.

What the 10-Q Reveals That the Prior Analysis Missed

The Samsung ICC award is a rate anchor with a fuse attached.

In July 2025, an ICC arbitral panel set total royalties of $1.05 billion for an eight-year patent license between InterDigital and Samsung covering Samsung products other than digital televisions and computer display monitors. The award covers cellular SEP patents in the smartphone context -- a different technology layer, different standards body, and different licensing framework from the video codec patents at issue in the Amazon dispute. It is not a directly applicable rate comparable. What it does signal is the scale of awards InterDigital is willing to pursue through arbitration, its demonstrated willingness to litigate to a large outcome rather than settle early, and a data point for board-level settlement authorization thresholds that likely influences how InterDigital calibrates its bilateral demands against Amazon.

The fuse: Samsung filed a request to challenge that award in December 2025. The outcome of that challenge is not externally observable. But if the ICC revises the award downward before September 2026, InterDigital’s internal rate floor weakens, its willingness to hold out for a high bilateral settlement figure diminishes, and the settlement range narrows in Amazon’s favor. This makes InterDigital more risk-averse about the UK trial than its aggressive multi-front posture implies, and that asymmetry is one of the two material changes to the settlement logic in Part 4.

Dolby escalated from PTAB to Article III federal court.

Part 3 identified the Dolby IPR petition against InterDigital US Patent 9,185,268 as a licensor coalition fracture. The 10-Q reveals something more immediate. In March 2026, Dolby filed two declaratory judgment complaints in the Central District of California. The first covers one patent also asserted against Disney. The second covers five patents also being asserted against Amazon, Hisense, and TCL.

Those five patents in Dolby’s Complaint 2 are the same patents being asserted against Amazon. Dolby is not just challenging InterDigital at the PTAB. It is seeking declaratory judgment in federal court on the exact patents at issue in the Amazon ITC investigation and the Eastern District of Virginia proceedings. InterDigital filed a motion to dismiss Complaint 1 in April 2026. The deadline to respond to Complaint 2 has not yet passed.

The analytical implication: if the California court denies InterDigital’s motion to dismiss and the case proceeds, discovery and claim construction in California could produce adverse rulings on the Amazon-asserted patents before the ITC initial determination issues. That would weaken the ITC exclusion order threat that Part 3 identified as the key secondary driver. The Dolby threat is now more immediate and more direct than any prior analysis captured.

The German court proceedings are running on a 2026 trial schedule.

All three prior parts focused on the UPC as the primary European enforcement mechanism. The 10-Q reveals that InterDigital filed three separate infringement proceedings in Munich and Mannheim Regional Courts against Amazon in November and December 2025, and that trials are expected in two of those three proceedings in Q3 and Q4 2026.

German court infringement findings in Q3 or Q4 2026 would arrive before or alongside the September 2026 UK RAND trial. A Munich or Mannheim infringement finding creates injunction pressure on Amazon’s European device sales independent of the UPC AILI proceedings. The prior analysis treated European enforcement pressure as flowing exclusively through the UPC. That was incomplete.

The DOJ filed a Statement of Interest in the Disney Delaware antitrust case -- and it favored InterDigital.

In August 2025, Disney filed an antitrust complaint against InterDigital and Technicolor in Delaware alleging monopolistic conduct in video codec SEP licensing. In October 2025, the Antitrust Division of the United States Department of Justice filed a Statement of Interest in that case.

The prior analysis framed this as regulatory pressure against InterDigital. That framing was incorrect. The DOJ’s Statement of Interest steered the dispute away from antitrust law, consistent with the Trump administration’s broader posture of limiting antitrust scrutiny of SEP licensing disputes. InterDigital filed a motion to dismiss Disney’s antitrust claims. The DOJ’s filing was characterized by observers as tilting toward InterDigital rather than constraining it.

The correct analytical implication is the opposite of what was previously written. DOJ involvement in this posture reduces the regulatory flanking risk to InterDigital in the US rather than increasing it. It signals that the federal government is unlikely to treat InterDigital’s SEP licensing conduct as an antitrust violation, which strengthens InterDigital’s enforcement posture rather than constraining it.

The Revised Pressure Map

The proceedings map has changed significantly from Part 1. Here is the complete picture as of April 30, 2026:

Against Amazon:

The UK High Court proceedings are active, with jurisdiction confirmed in December 2025 and a RAND trial scheduled for September 2026. The ITC investigation (337-TA filed December 2025) has been formally instituted. The Delaware companion case is stayed pending ITC resolution. The Eastern District of Virginia case (four video coding patents, filed December 2025) is active. The Western District of Texas case (six Edgio-acquired content delivery patents, filed February 2026) is active and targets Amazon’s cloud infrastructure rather than its devices. Three separate German court proceedings are expected to reach trial in Q3 and Q4 2026. Three separate UPC proceedings are active across Mannheim and Dusseldorf Local Divisions.

Against other defendants:

InterDigital obtained a preliminary injunction against Transsion in Brazil in March 2026, enjoining Transsion from selling 5G-compliant devices. The Disney UPC proceedings have hearings scheduled in June and July 2026 in Dusseldorf, which will produce infringement findings before the September 2026 Amazon UK trial. Munich imposed fines of 550,000 euros on Disney in January 2026 for violations of an injunction. The Hisense and TCL Munich proceedings have hearings scheduled for January 2027.

The Lenovo arbitration adds a second ongoing rate proceeding.

InterDigital and Lenovo agreed to binding ICC arbitration in Q4 2024 to determine the final terms of a new patent license effective from January 1, 2024. The ICC confirmed the full tribunal in March 2025. The arbitration hearing is expected before year end 2026. Like the Samsung award, the Lenovo arbitration primarily covers cellular SEP patents rather than video codec patents, so it is not a direct rate comparable for the Amazon proceedings. Its significance is the timing: an ICC arbitration outcome expected before year end 2026 creates a second contemporaneous licensing data point that may influence how InterDigital’s internal rate expectations are set and how its litigation posture is calibrated heading into the September 2026 UK trial.

The AV1 Dimension: Not a Threat to InterDigital, an Expansion of Its Arsenal

The Part 4 swarm surfaced AV1 as a latent factor, modeling it as a risk to InterDigital’s portfolio on the assumption that AV1 is royalty-free and gaining streaming market share. That framing was wrong and the article previously published it incorrectly. The correction matters.

InterDigital holds AV1 patents and is actively asserting them against Amazon. The Delaware complaint defines “AV1/VP9 Accused Instrumentalities” as all Amazon devices capable of decoding AV1 or VP9 content, including Fire Stick 4K, Fire TV Stick 4K Max, Fire TV Cube, Fire Max 11, Echo Show 15, and Echo Show 21. US district court lawsuits over AV1 reached an all-time high in 2025, with InterDigital among the companies driving that increase.

The correct analytical framing is the opposite of what was previously written. AV1 adoption by Amazon devices does not depreciate InterDigital’s enforcement position -- it expands it. Every Amazon device that implements AV1 decoding becomes an additional accused instrumentality. The “royalty-free” characterization of AV1 reflects the Alliance for Open Media’s aspiration for the codec, not a legal determination that no patents cover its implementation. InterDigital and others have asserted that their patents are fundamental to AV1 implementation regardless of the consortium’s royalty-free design intent.

Amazon is a founding member of the Alliance for Open Media, the consortium that developed AV1. The irony is that Amazon’s own investment in developing an allegedly royalty-free alternative to HEVC has not insulated it from patent assertions on that very codec. If anything, AV1 adoption increases the scope of the dispute rather than constraining it.

This correction removes the AV1 adoption risk as a settlement pressure variable on InterDigital. It replaces it with the opposite: AV1 as an additional enforcement layer that expands Amazon’s total exposure beyond the HEVC/HDR patents originally identified.

The Revised House View

Prediction: Amazon and InterDigital conclude a bilateral global patent license agreement in Q3 or Q4 2026, before the UK High Court issues a formal binding RAND judgment. The settlement is structured as a per-unit running royalty on FRAND-committed video coding patents (HEVC/AVC/VVC) with a lump-sum back-payment, plus a separate commercial arrangement for non-FRAND-committed patents including the Edgio-acquired content delivery portfolio, which carries no RAND constraint and therefore gives InterDigital uncapped leverage on that technology layer. The ITC investigation and UPC proceedings are withdrawn or stayed by consent. The UK proceedings are stayed rather than dismissed, preserving the rate-anchoring function without producing a binding precedential judgment.

Alternative Outcome: The UK High Court issues a binding global RAND rate determination in September or October 2026. This occurs if the UPC Court of Appeal narrows the AILI on May 28, 2026, reducing Amazon’s injunction exposure sufficiently that it rationally holds out for a court-determined rate; the ITC timeline slips into 2027, removing peak-demand import ban pressure; and the Samsung ICC award is revised downward on challenge before September 2026, weakening InterDigital’s rate floor and making its bilateral demands appear inflated relative to the emerging comparable license landscape. Amazon accepts the court-determined rate as the license basis within 60 to 90 days of judgment. The Edgio content delivery patents remain in active litigation in Texas as a separate track. Probability: approximately 30 to 35 percent.

Key Driver: The outcome of the May 28, 2026 UPC Court of Appeal hearing on the AILI scope, in combination with the ITC initial determination timeline. These two events, both resolving before the September 2026 UK trial, jointly determine whether Amazon’s multi-jurisdictional injunction exposure is severe enough during the Q4 2026 holiday cycle to compel pre-judgment settlement, or whether Amazon retains sufficient optionality to hold out for the court-determined rate.

Secondary Driver: The Samsung ICC award challenge. The award covers cellular SEP patents rather than video codec patents and is not a direct rate comparable for the Amazon proceedings. But if the challenge succeeds and the award is revised downward before September 2026, it may affect InterDigital’s internal settlement posture and its board-level authorization threshold for bilateral deals -- signaling reduced willingness to hold out for large litigation outcomes across its licensing program generally. This dynamic is entirely invisible in court filings and turns on ICC arbitral proceedings whose timeline and outcome are not externally observable.

Main Uncertainty: Whether the Dolby Complaint 2 California DJ action produces adverse claim construction on the Amazon-asserted patents before the ITC initial determination. If it does, the ITC exclusion order threat weakens materially and Amazon’s calculus shifts toward holding out for the UK judicial rate.

Confidence: 61 percent.

What Changed Across Four Runs

The directional prediction has held across all four iterations: confidential global license, not a public judgment. The mechanism, timing, and confidence have evolved with each run.

Part 1 predicted Q4 2026, driven by the 50 million euro UPC penalty. Confidence: 52 percent.

Part 2 corrected the Delaware error, added the full US docket picture, and moved the window to late 2026 or early Q1 2027. Confidence: 54 percent.

Part 3 added the validity challenge layer, the regulatory flank, and the ITU Common Patent Policy analysis. The window sharpened to July or August 2026. Confidence: 68 percent.

Part 4, anchored to the SEC-filed 10-Q, adds the Samsung ICC rate proceeding, the Dolby California DJ action, the German court trial schedule, the corrected DOJ framing (which favors InterDigital rather than constraining it), and the AV1 correction (InterDigital asserts AV1 patents against Amazon rather than facing AV1 as a portfolio risk). The window widened slightly to Q3 or Q4 2026 to reflect the additional complexity of the pressure map. Confidence settled at 61 percent, reflecting both the richer fact base and the genuine uncertainty introduced by the Samsung award challenge and the Dolby proceedings.

The most important insight across the entire series remains the Adversarial Critic’s contribution from Part 3: InterDigital’s private projection of the UK RAND rate is the unobservable variable that could override the convergence logic regardless of what the courts do. The 10-Q does not resolve that uncertainty. It adds the Samsung ICC award as a signal of InterDigital’s willingness to litigate to large outcomes -- while correctly noting it covers cellular SEP patents rather than video codec patents and is not a direct rate comparable -- and it adds the ICC challenge as a mechanism that could soften InterDigital’s litigation posture if the award is revised downward.

A Note on Methodology

Part 4 introduces one new data source: the InterDigital Q1 2026 10-Q litigation section, obtained from the SEC EDGAR database and processed as the primary seed document. No external API calls were required for this run. The 10-Q text was read directly.

The total input file for Part 4 was 54,808 characters, the largest of the series. A 5-second rate limiting delay was added between API calls to manage the larger token volume per request. The full pipeline, 57 API calls across ten agents, five rounds, coverage audit, and final refinement, cost approximately $4.00 at current Claude Sonnet 4.6 pricing. Runtime was approximately 20 minutes.

The full Python implementation including the 10-Q pre-processor is available at github.com/theharlans/standards-at-risk.

Sources: InterDigital Inc. Form 10-Q for the quarterly period ended March 31, 2026, filed April 30, 2026 (SEC EDGAR). JUVE Patent, “InterDigital vs Amazon: A chronology of the escalation,” March 16, 2026. CourtListener/PACER docket data retrieved April 2026. All analysis is the author’s own. This is not legal advice.

The short version

• The EU, Japan, and China have each issued significant IP licensing instruments within the past six weeks, acting in parallel but without coordination.

• Their vectors point in different directions: the revised EU Technology Transfer Block Exemption Regulation tilts toward licensor protection by eliminating the Licensing Negotiation Group safe harbor; Japan’s new draft guidelines target dominant-buyer extraction of IP and know-how from weaker suppliers; China’s post-DS611 architecture preserves anti-suit injunction capability while restricting the outbound transfer of data relevant to foreign licensing proceedings.

• Japan’s instrument contains an analytical gap worth naming directly: the same statutory framework that prohibits dominant buyers from extracting IP from smaller suppliers should, in principle, equally prohibit dominant implementers from engaging in holdout against smaller SEP licensors. The JFTC has not addressed that symmetry, and it should.

• For SEP practitioners and licensors, the integrated picture is neither uniformly favorable nor uniformly hostile. It is multi-polar. Strategy built around any single jurisdiction’s posture is increasingly incomplete.

I. Why this moment matters

Something significant happened between mid-March and late April 2026 that has received no coordinated attention: three of the world’s most consequential IP licensing jurisdictions each published major regulatory instruments affecting the environment in which SEP licensing negotiations occur. The European Commission issued its revised TTBER and Technology Transfer Guidelines on April 16. Japan’s JFTC, SMEA, and JPO jointly published draft guidelines on abuse of dominant position in IP, know-how, and data transactions on March 30. And China’s post-WTO DS611 regulatory architecture (covered in detail in April: “China Didn’t Retreat on SEPs. It Changed the Rules of the Game.” and “China’s Regulatory Upgrade: The Threat Nobody Is Talking About”) has continued to consolidate through implementation of State Council Decree No. 801 and Guoling No. 835.

None of these instruments was designed with the others in mind. They emerged from different domestic policy processes, different institutional cultures, and different industrial constituencies. But they are landing simultaneously on the same multinational licensors and implementers, and their combined effect is something more complex and more consequential than any individual instrument suggests.

II. The EU: a genuine pro-licensor signal

I covered the TTBER in detail when it published in April (“The EU’s New Tech Transfer Guidelines”), so I will not re-argue it here. The headline for SEP licensors was the removal of the Licensing Negotiation Group safe harbor. LNGs (collective implementer negotiation groups) had enjoyed safe harbor treatment under the prior framework on the theory that they helped balance licensor bargaining power. The revised regulation eliminates that protection. The Commission’s reasoning, which I find correct, is that LNGs in practice function as coordinated holdout mechanisms rather than efficiency-promoting structures. Removing the safe harbor is an acknowledgment that implementer coordination is a real competition problem, not a theoretical one.

That is a meaningful pro-licensor move. The revised TTBER and guidelines still leave important questions unresolved, particularly around SEP pool analysis and the treatment of portfolio licensing. But the directional signal is clear, and it matters because the EU is, for now, still the jurisdiction in which the foundational FRAND precedents were established and in which the most technically sophisticated SEP litigation occurs.

III. Japan: the mirror-image problem

Japan’s March 30 draft instrument is less visible outside Japanese practitioner circles, and its connection to SEP licensing is indirect. It deserves more attention than it has received.

The backstory: the JFTC has been building toward this instrument since a major 2019 fact-finding survey documented systematic extraction of patents, know-how, and technical data from smaller Japanese manufacturing suppliers by dominant buyers (large OEMs and system integrators who leveraged purchasing relationships to obtain IP at below-market or zero compensation). A new “IP Transaction Normalization Working Group,” chaired by attorney Izumi Hayashi of Sakurazaka Law Office, ran from August 2025 through March 2026, producing a report that the draft guidelines directly implement. The joint JFTC/SMEA/JPO instrument addresses abuse of superior bargaining position in transactions involving IP rights, know-how, and data (a meaningful new addition to the prior framework’s scope).

The target is the dominant buyer who coerces a weaker supplier into disclosing trade secrets without fair compensation, assigning patents as a condition of continued business, tolerating unilateral IP ownership terms in joint development contracts, or providing manufacturing process data without adequate consideration. Those are genuine, well-documented problems in Japanese supply chains, and the instrument is a serious response to them.

The statute that governs this conduct, however, is Article 2(9)(5) of the Antimonopoly Act, which prohibits abuse of superior bargaining position. It is not directionally limited. It applies wherever one party holds superior bargaining position over another and exploits it to impose unjust disadvantage. The draft guidelines address only one polarity of that relationship.

The mirror scenario is analytically identical. A dominant implementer (a major automotive OEM, a large consumer electronics platform, a global device manufacturer) who refuses to engage in good-faith SEP licensing negotiations with a smaller licensor is exploiting superior bargaining position to impose unjust disadvantage on a party who cannot easily walk away. The licensor who has made a FRAND commitment cannot simply refuse to deal. The implementer who knows that knows it. Sustained holdout by a dominant implementer is, under the same framework the JFTC is now developing, structurally indistinguishable from the extraction conduct the guidelines are designed to prevent.

The JFTC’s existing IP Guidelines, last amended in 2016, do address SEP-specific issues including injunctions and refusal to license. But they operate within an exclusionary-conduct framework, not the relational abuse-of-position framework that the new instrument develops. That is a different analytical register, and the gap between them is real. The March 2026 draft does not bridge it.

This matters practically, not just analytically. Japan’s automotive and IoT sectors (among the primary battlegrounds for next-generation SEP licensing in 5G and 6G applications) are precisely where dominant-implementer holdout dynamics are most pronounced. Many Japanese component suppliers hold declared SEPs. Many of their largest customers are the same OEMs whose supply-chain IP practices the new guidelines are designed to constrain. The JFTC has, in effect, developed a sophisticated analytical framework for one side of that relationship. The question it has not answered is whether that framework applies symmetrically.

I do not think the JFTC made a deliberate decision to exclude SEP licensing relationships from the instrument’s scope. The more likely explanation is that the Working Group’s mandate was framed around supply-chain normalization, and SEP licensing simply was not in view. But the logical implication of the instrument, read carefully, is that dominant-implementer holdout should be as analytically suspect as dominant-buyer IP extraction. The JFTC should say so explicitly in the final version of the guidelines.

The public comment window for the March 30 draft closed today, April 28. I assessed whether to file a formal comment, but the submission requirement is Japanese-language only and the timeline (48 hours after I became aware of the instrument) made a properly drafted submission impractical. This article is what I would have submitted. There will be a final instrument, likely later in 2026, and that will be the right moment to press this argument on the record. I intend to do so.

IV. China: recalibration, not retreat

My April coverage traced China’s post-DS611 regulatory architecture in detail (“China Didn’t Retreat on SEPs. It Changed the Rules of the Game.” and “China’s Regulatory Upgrade: The Threat Nobody Is Talking About”), and I will not reconstruct the full argument here. The short version is that China adjusted its formal procedural posture in ways minimally sufficient to address the WTO dispute panel’s findings while preserving the substantive tools (ASI capability, domestic jurisdiction preference, data transfer restrictions) that make Chinese proceedings operationally significant in global SEP licensing strategy.

The provision I continue to watch most closely is Article 8(4) of the DS611 implementing rules, which restricts the outbound transfer of certain data and evidence in the context of SEP licensing disputes. A licensor building a global evidentiary record for a FRAND rate determination (the kind of proceeding that requires comparable license data, technical contribution analysis, and portfolio valuation) faces structural friction when the relevant commercial and technical data sits with Chinese implementers or in Chinese corporate systems. That friction is not incidental. It is a designed feature of a regulatory architecture whose overall orientation is implementer-favorable.

V. The integrated picture

A licensor operating across all three jurisdictions simultaneously now faces this environment: a European framework that has signaled genuine support for licensor interests by removing LNG protection, but that still leaves SEP pool treatment and portfolio licensing questions incompletely resolved; a Japanese framework that has developed sophisticated tools for addressing IP exploitation in vertical relationships, but that has not yet applied those tools symmetrically to SEP holdout dynamics; and a Chinese framework that presents a compliant face to WTO scrutiny while preserving the architecture that makes domestic proceedings attractive to implementers.

The net effect is not uniformly favorable or hostile. It is multi-polar. And the absence of any coordination mechanism among these three frameworks means that jurisdictional arbitrage (by both licensors and implementers) will continue to drive where disputes are initiated and where they resolve.

That dynamic will not improve on its own. The instruments themselves are not moving toward convergence. The standards community, and specifically those practitioners and organizations with the standing to engage each of these frameworks at the regulatory level, should be doing so actively, in all three jurisdictions, simultaneously. Engaging one at a time is no longer adequate.

Jim Harlan is an independent intellectual property practitioner with approximately 25 years of experience in SEPs, FRAND licensing, and standards policy. Standards at Risk covers patent licensing, global standards policy, and innovation market dynamics. Subscribe at standardsatrisk.substack.com.

Why Tesla Filed in the UK and Why It Matters

In December 2023, Tesla filed proceedings in the UK High Court seeking two things: declarations that several InterDigital patents are invalid or non-essential, and a court determination of FRAND terms for a license to the entire Avanci 5G Vehicle platform.

The “why UK” question has a plausible answer once you understand the procedural logic. Under the framework established by Unwired Planet v. Huawei and affirmed in subsequent cases, UK courts will set global FRAND rates for portfolios that include UK patents. Tesla may have chosen InterDigital as the anchor defendant in part because InterDigital holds UK patents, giving the court a potential jurisdictional hook. Tesla then attempted to use InterDigital as a representative defendant under CPR rule 19.8 to bring all 65-plus Avanci pool members within the court’s reach in a single proceeding. One hearing, one rate, covering approximately 11,900 UK SEPs across the entire connected vehicle pool.

Tesla’s own filings suggest the motivation. It says it was pressured into taking a 4G Avanci license by litigation from pool members seeking injunctions that threatened to bar its 4G-enabled vehicles from important markets. It filed in the UK for 5G more than a year before its 5G vehicles were ready to deploy, suggesting a pre-emptive posture aimed at getting a court-determined rate before similar pressure could be applied again.

The rate in dispute is $32 per vehicle, which is Avanci’s flat-rate offer for a license to the 5G platform. Tesla contends this exceeds a FRAND rate and was presented on a non-negotiable basis. Avanci’s response is that the rate reflects a transparent, one-stop-shop structure that more than 150 automotive brands worldwide have accepted across its 4G and 5G programs, and that Tesla is not being asked to pay anything different from what every other vehicle manufacturer pays.

What the Courts Below Found

Both the High Court and the Court of Appeal declined to permit Tesla’s FRAND determination claim against InterDigital as a representative defendant for all Avanci pool members. The procedural mechanism extends the court’s reach from the patents actually before it to an entire pool built around 65-plus independent companies, the vast majority of whom never submitted to UK jurisdiction at all.

The Avanci Case: A Pool That Works Is Not a Problem to Be Solved

The numbers here are worth stating plainly, because they are the most important context for understanding what is actually at stake in this hearing.

Avanci’s 4G and 5G Vehicle programs together cover more than 275 million connected vehicles worldwide. More than 150 automotive brands participate across both programs. The 5G program alone involves approximately 90 licensors and has brought in Toyota, General Motors, Volkswagen, Ford, Volvo Cars, Honda, Jaguar Land Rover, Nissan, and most recently seven Chinese automakers in agreements announced in March 2026. The Court of Appeal’s own record notes the contrast between the “orderly licensing” Avanci achieved in the automotive sector and the “licensing debacle” in the mobile phone sector where no comparable pool exists.

This is a pool that is working. The automotive industry has accepted it overwhelmingly. The rate structure is transparent, non-discriminatory, and applied equally to every licensee from the largest European manufacturer to the newest Chinese entrant. The pool has even extended its model to Wi-Fi 6, launching a new program in March 2026 with Mercedes-Benz as the inaugural licensee.

Tesla’s position is that the $32 per vehicle rate is too high and was offered on a take-it-or-leave-it basis. Those are legitimate commercial concerns. But the mechanism Tesla chose to address them raises a question about whether it constitutes a FRAND compliance remedy or something closer to a litigation strategy that, if successful, could give any automotive manufacturer the option of seeking judicial repricing of a pool that the rest of the industry has accepted through negotiation. The mechanism involves UK litigation seeking a court-imposed rate covering all 65-plus pool members through a procedural device that bypasses their individual consent to jurisdiction.

The downstream effect on the pool model is worth considering as an under-discussed consequence of a Tesla win at the UKSC. If a single implementer can anchor UK jurisdiction to one pool member and then seek a globally binding rate covering all pool participants, the incentive for other implementers to negotiate bilaterally may be diminished. The bilateral alternative, meaning individual negotiation with each of 65-plus patent holders, is precisely what Avanci was built to avoid, and what Avanci’s own counsel described as the “licensing debacle” in mobile. A UKSC decision that enables the Tesla model could make the pool model more legally fragile at exactly the moment when it is demonstrably succeeding.

Why InterDigital and Why December 2023

One question worth addressing directly is why Tesla may have chosen InterDigital as the anchor defendant and why it filed when it did.

The timing is not coincidental. In March 2023 the UK High Court issued its first instance decision in InterDigital v. Lenovo, setting a global FRAND rate of $0.225 per device for InterDigital’s cellular SEP portfolio. That figure was ultimately ~65% below InterDigital’s ask and far closer to Lenovo’s position. The High Court also found that InterDigital had not acted as a willing licensor, having consistently sought supra-FRAND rates during negotiations. Tesla filed its UK proceedings in December 2023, nine months after that decision entered the public record.

The timing invites a plausible inference. InterDigital was arguably the most prominent Avanci pool member against whom a UK FRAND rate had been publicly determined. Tesla’s apparent strategy may have been to use InterDigital as a jurisdictional anchor, with its UK patents giving the court a potential hook, and then argue that a judicially determined automotive rate would be materially lower than Avanci’s $32 flat offer, using the Lenovo cellular rate as a possible reference point. One pool member, one UK filing, one rate covering the entire platform.

The problem with that strategy, beyond the representative defendant jurisdictional question that defeated it at first and second instance, is that the Lenovo rate is a poor comparator for the Avanci automotive context for reasons the subsequent record makes clear.

The Court of Appeal in July 2024 corrected the first instance rate materially upward to $0.30 per device and increased the total lump sum Lenovo must pay to over $240 million, roughly three times Lenovo’s original offer. InterDigital itself described the Court of Appeal outcome as a resounding victory. Both the UK High Court in subsequent proceedings and the Munich Regional Court explicitly cautioned against using the Lenovo decision as a forward-looking rate reference. Justice Reynolds noted he did not “have a high degree of assurance that this is a FRAND rate” for any future license, and Munich observed that at least a 30% inflation uplift would be mandatory for any forward-looking calculation.

There is a deeper structural problem with Tesla’s comparator argument. The Lenovo decision covered a cellular SEP portfolio licensed to handset manufacturers, a market with decades of licensing history, established comparable transactions, and well-developed royalty benchmarks. The Avanci 5G Vehicle platform covers connected vehicle technology across a pool of approximately 90 licensors offering a transparent, non-discriminatory flat rate to an automotive industry that had never previously navigated multi-licensor cellular SEP licensing at scale. The technology layer, the licensing context, the implementer base, and the pool structure are all materially different. A cellular handset rate is not a comparable for a connected vehicle pool rate, and the courts recognized that Tesla was attempting to use one rate-setting exercise as a lever to reach a completely different licensing environment.

One reasonable reading of Tesla’s strategy is that it attempted to use a single backward-looking rate determination, one that was subsequently corrected on appeal, as a template for seeking judicial pricing on a pool that 150-plus automotive brands had already accepted through negotiation. The UK courts declined to allow that mechanism to succeed at either instance level. The UKSC will now decide whether the jurisdictional question that enabled the attempt in the first place should remain available to implementers in future pool disputes.

The Jurisdictional Escalation Context

This hearing does not occur in isolation. It occurs against a background of accelerating jurisdictional conflict over who controls the global licensing environment for standard-essential technology, a conflict this publication has analyzed in prior Standards at Risk articles on China’s regulatory response to WTO dispute DS611.

The relevant context is this: the UK’s extraterritorial FRAND rate-setting ambition, established in Unwired Planet and developed through the Optis and InterDigital v. Amazon cases, has prompted regulatory counter-moves in other jurisdictions. China’s courts began setting their own global FRAND rates. That led to anti-suit injunctions, counter-injunctions, and eventually the Munich Regional Court’s unprecedented pre-emptive anti-interference order. The UPC has issued its own ASIs affecting UK proceedings. The European Commission was formally notified of one such order in December 2025.

The possibility of a WTO complaint against the UK over its FRAND rate-setting practice has been raised in commentary on this hearing. That risk may not be trivial. The EU’s DS611 complaint against China was premised on China’s use of ASIs to suppress royalty rates for cellular SEPs. A structural argument could potentially be developed that the UK’s extraterritorial rate-setting constitutes an analogous distortion, allowing implementers to forum-shop for a favorable judicial determination. Whether such an argument would succeed is a separate question, but a jurisdiction representing approximately 3% of global GDP that is setting licensing terms for technology deployed worldwide may face legitimate questions about the international standing of that practice.

The pro-licensor reading of this dynamic is not that UK courts should stop setting FRAND rates. It is that the framework needs to be stable, internationally recognized, and limited to cases where UK patents are genuinely central, rather than cases where UK patents are used as a procedural hook to reach an entire pool most of whose members have no UK connection.

What a Durable Outcome Looks Like

The UKSC is not being asked to overturn Unwired Planet. It is being asked to define the outer boundary of the representative defendant mechanism in the pool context. That is an opportunity to improve the framework rather than simply affirm or reverse it.

A durable outcome from a pro-licensor and pro-pool perspective might do three things.

First, it would confirm that UK courts can set FRAND rates for patents actually litigated before them. That is settled law and should remain so. The Avanci model benefits from a predictable judicial backstop. Licensees who refuse bilateral negotiation can be brought before a court, and the existence of that option is part of what makes the pool’s take-it-or-leave-it offer credible.

Second, it would decline to extend the representative defendant mechanism to pool members who have not individually submitted to UK jurisdiction. The pool structure works because participation is voluntary and the rate is transparent. Allowing one member to be used as a procedural proxy for all others removes the individual consent that makes pool governance legitimate. This is not a limitation on UK FRAND jurisdiction. It is a recognition that jurisdiction follows the patents, not the pool.

Third, it would do so in terms that are internationally legible, acknowledging the jurisdictional escalation dynamic and signaling that UK courts are not seeking to set rates for technology whose primary markets lie elsewhere. That signal matters not just for future litigation but for the diplomatic and regulatory environment in which UK courts operate.

Tesla has a legitimate interest in paying a FRAND rate. Avanci has a legitimate interest in maintaining a pool structure that has delivered orderly licensing to more than 275 million connected vehicles. Both interests can be served by a framework that gives Tesla access to UK patent validity and essentiality determinations, which the courts below did not foreclose, while declining to give it a mechanism to impose a judicially determined rate on 65-plus independent companies that never consented to UK jurisdiction.

The UKSC has an opportunity to make the Unwired Planet framework more durable by defining its limits with precision. A maximalist affirmance of the representative defendant mechanism could extend UK reach in ways that accelerate the jurisdictional conflict it has already generated, while potentially making the pool model more legally fragile at exactly the moment when it is demonstrably succeeding.

The Broader Stakes for Patent Pools

The Avanci story is worth dwelling on because it is the best empirical evidence available for what a well-functioning SEP licensing structure looks like in a complex multi-technology, multi-licensor environment.

More than 275 million connected vehicles. More than 150 automotive brands. Approximately 90 licensors in the 5G program. Seven Chinese automakers joining in March 2026 alone. A new Wi-Fi 6 program launched with Mercedes-Benz. The contrast with the mobile phone sector’s bilateral licensing friction, which produced years of multi-jurisdictional litigation involving Apple, Huawei, Samsung, Ericsson, Nokia, and others, is exactly as stark as Avanci’s counsel described it.

Patent pools work when they offer transparent, non-discriminatory rates that reduce transaction costs for everyone. Avanci has done this in the automotive sector. The question before the UKSC is whether the pool model is protected by a legal framework that respects its structure, or whether any single holdout can use judicial machinery to reprice what the rest of the industry has negotiated and accepted.

That question matters well beyond Tesla and Avanci. The connected vehicle sector is arguably the first major proving ground for SEP licensing in non-handset contexts. The IoT sector, the industrial automation sector, and the smart infrastructure sector all face similar challenges: how do you efficiently license standard-essential technology across thousands of implementers without the bilateral licensing friction that characterized the mobile era? Patent pools represent one market-developed answer. The UKSC hearing may well reinforce or undermine the legal foundation on which that answer rests.

The UK Supreme Court hearing in Tesla v. InterDigital and Avanci runs from 28 to 30 April 2026. A decision is expected in the months following. Standards at Risk will analyze the ruling when it issues.

Sources: UK Supreme Court case record UKSC/2025/0058; Court of Appeal judgment [2025] EWCA Civ 193; High Court judgment [2024] EWHC 1815 (Ch); Avanci Vehicle program announcements March 2026; ICLE intervention brief March 2026; Lawyer Monthly, March 2026. All analysis is the author’s own. This is not legal advice.

• The European Commission has released its first comprehensive update to the Technology Transfer Guidelines since 2014. The core TTBER architecture is unchanged, but three developments matter for the SEP and FRAND world.

• The LNG safe harbor removal is a licensor win. Without a safe harbor, every joint implementer negotiation arrangement faces individual assessment and real Article 101(1) exposure from inception. Some commentary has this backwards.

• The ACCC Apple Pay case from 2017 is the best available precedent for understanding what competition authorities actually do when implementer coalitions formalize collective negotiation and boycott strategies. Australia’s regulator denied four major banks authorization to collectively bargain with and boycott Apple, finding the harm ran not just to Apple but to competitive tension among the banks themselves. The reasoning maps directly onto the EU LNG framework.

• Paragraph 322 is the most practically significant development for SEP licensors. LNG members engaging in coordinated negotiating delays may be prevented from invoking the Huawei v ZTE FRAND defense against injunctions. This is a structural deterrent to holdout dressed as collective negotiation.

• The guidelines contain a significant analytical gap on SEP pools. Generic pool skepticism is applied without distinguishing optional SEP pools, where every contributor retains a bilateral FRAND obligation, from non-SEP aggregations that can actually exercise market power. That gap will be exploited in litigation.

• These are soft law. The guidelines do not bind national courts and do not preclude private enforcement, but they function as interpretive reference points in European litigation, arbitration, and FRAND disputes globally.

Introduction

The European Commission published its revised Guidelines on the application of Article 101 TFEU to technology transfer agreements on April 16, 2026, the first comprehensive update since 2014 (https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52026XC04/16). The accompanying Technology Transfer Block Exemption Regulation enters into force on May 1, 2026.

A necessary framing point: these are guidelines, not binding law. They do not bind national courts, do not preclude private enforcement, and do not immunize conduct from judicial review. What they do is signal DG COMP’s enforcement priorities and provide interpretive reference points that courts and arbitral tribunals routinely consult. That is consequential, but it is not the same as a regulation, and overclaiming in either direction does the analysis a disservice.

The core TTBER architecture is unchanged. Market share thresholds stay at 20% for competitors and 30% for non-competitors. The hardcore and excluded restrictions are substantively the same. The guidelines reaffirm that the great majority of technology licensing agreements are pro-competitive and that innovators should be free to seek appropriate remuneration, accounting for ex ante investment and risk. For a pro-licensor practitioner, those are not throwaway lines. They are the doctrinal foundation on which individual FRAND disputes continue to be argued.

Three developments in the new guidelines deserve close attention. The rest is incremental.

Issue One: The LNG Safe Harbor Removal Is a Licensor Win

The LNG section is entirely new to the guidelines. Licensing negotiation groups, arrangements whereby implementers jointly negotiate technology license terms, received no safe harbor. Some commentary has characterized this as the Commission remaining favorable to LNGs or as a backtrack from the draft. That reading is backwards.

A safe harbor is a bright-line zone of legal certainty. Its absence means every LNG arrangement, regardless of how carefully structured, faces individual assessment, requires documented compliance analysis, and carries real Article 101(1) exposure. That is a structural deterrent to LNG formation, not an accommodation of it. Practitioners advising implementer groups will need to conduct that individual assessment carefully and document it thoroughly. There is no comfort zone.

The ACCC Apple Pay Case Shows What This Looks Like in Practice

The concern about collective implementer negotiation tipping into buyer cartel conduct is not theoretical. In 2017, Australia’s competition regulator decided exactly this question, and the reasoning maps directly onto the EU LNG framework.

In July 2016, four major Australian banks, Commonwealth Bank, Westpac, NAB, and Bendigo and Adelaide Bank, applied to the Australian Competition and Consumer Commission for authorization to collectively bargain with Apple over Apple Pay access terms, specifically NFC chip access on iPhones. They also sought authorization for a collective boycott of Apple Pay while negotiations proceeded. Under Australian law, such collective conduct requires explicit regulatory authorization to avoid cartel liability (https://www.accc.gov.au/media-release/accc-denies-authorisation-for-banks-to-collectively-bargain-with-apple-and-boycott-apple-pay).

The ACCC denied both requests in its final March 2017 determination. Chairman Rod Sims stated the authority was “not satisfied, on balance, that the likely benefits from the proposed conduct outweigh the likely detriments.” Apple’s submission was direct: allowing competing firms to form a collective to dictate terms to a technology provider would set a troubling precedent and delay the introduction of disruptive technologies.

The critical finding was not just harm to Apple. The ACCC found that collective bargaining would reduce competitive tension between the banks themselves in their individual negotiations, and distort downstream competition in mobile payments markets. When every bank negotiates as a bloc, Apple cannot play one bank’s terms against another’s, and the banks cannot differentiate their services through the terms they secure. The collective arrangement that appears to strengthen implementer bargaining power against the licensor simultaneously weakens competitive discipline among the implementers.

The banks narrowed their application in early 2017, dropping the fee dispute and focusing solely on NFC access, and halving the proposed authorization period to 18 months (https://www.reuters.com/article/us-apple-australia-banks/australian-banks-narrow-focus-of-apple-pay-collective-bargaining-request-idUSKBN15S0DX). The ACCC was not persuaded. The final denial stood.

The parallel to SEP LNGs is direct. A group of downstream competitors pooled their negotiating leverage against a technology gatekeeper, combined collective negotiation with a threatened collective boycott, and needed regulatory authorization precisely because the conduct would otherwise have attracted cartel liability. The ACCC identified harm not only to the technology counterparty but to downstream competition among the implementers, which is exactly the cost commonality and downstream coordination concern in paragraph 326 of the new EU guidelines.

The technology context differs. The Australian banks faced no FRAND equivalent and Apple Pay NFC is not a declared-essential standard patent. The ACCC case is not direct authority in the SEP/LNG context. But it shows how competition authorities actually reason about collective implementer negotiation when the implementer-side efficiency arguments are tested against a real set of facts. The EU guidelines operate on the same logic, dressed in effects-based analysis rather than a flat prohibition.

The collective boycott element also deserves note. The banks’ request to withhold Apple Pay adoption during negotiations is structurally equivalent to coordinated holdout in SEP licensing. The ACCC treated it as a distinct and serious harm. The EU guidelines’ paragraph 322 does the same, as the next section addresses.

The Internal Tension the Guidelines Do Not Resolve

There is an asymmetry in the guidelines worth flagging for licensor-side counsel. The pool section warns that patent holders inside a pool may lack incentive to license bilaterally because doing so would undermine the pool’s collective pricing. The same incentive logic applies on the implementer side. An LNG whose members have strong commercial incentives to direct all licensing interactions through the group, and resist bilateral engagement with patent holders, poses the same structural concern. The guidelines acknowledge coordinated implementer coercion in paragraph 321 but treat it as a residual issue rather than a structural one. That gap is worth surfacing in any enforcement or litigation context.

Issue Two: Paragraph 322 and the Holdout Signal

Paragraph 322 is the most practically significant development in the document for SEP licensors, and it has not received the attention it deserves.

The paragraph addresses coordinated delays during LNG negotiations. Where delays are temporary and objectively linked to the pursuit of joint negotiations, the Commission will assess them as part of the overall LNG analysis. But the paragraph immediately adds that LNG members engaging in delays may be prevented from invoking Article 102 as a defense to a claim for a prohibitory injunction by a standard-essential technology holder that has committed to grant licenses on FRAND terms. The citation is Huawei Technologies v ZTE, C-170/13, ECLI:EU:C:2015:477 (https://curia.europa.eu/juris/document/document.jsf?docid=165911&doclang=EN).

This is the Commission telling implementers that joining an LNG and running a coordinated delay strategy does not insulate them from injunctive relief. The Huawei v ZTE framework established that an SEP holder who has made a FRAND commitment and made a FRAND offer can seek an injunction against an implementer who delays or obstructs. Paragraph 322 extends this to the LNG context: collective negotiating delay does not transform the underlying dynamic in a way that allows LNG members to invoke Article 102 as a shield against injunctions.

For licensors holding FRAND-committed SEPs who face organized implementer groups engaged in extended negotiating delays, this paragraph provides meaningful doctrinal support. It should be in every SEP litigator’s briefing file. Licensors should document the timeline, nature, and coordination patterns of any collective delay behavior by implementer groups from the earliest stages of any licensing engagement.

Issue Three: The SEP Pool Analytical Gap

The pool safe harbor at paragraph 286 survives with its eight conditions, but the guidelines apply generic pool skepticism in a way that creates a significant problem for SEP pool administrators.

The Commission’s concern about pools, including the paragraph 280 warning that pool members may have “little incentive to license independently in order not to undermine the licensing activity of the pool,” is analytically coherent in the context of non-SEP technology aggregations. If a pool bundles substitute technologies and licensees have no practical alternative, the pool can exercise real market power.

SEP pools are structurally different. Every SEP holder contributing to a pool retains a bilateral FRAND licensing obligation. A licensee who dislikes the pool’s terms can negotiate directly with individual patent holders, each of whom is bound to offer FRAND terms. The pool competes against bilateral alternatives constrained by the same FRAND commitment that justifies the pool’s existence. In this environment, the generic market power concern does not hold in the same way, and the guidelines should have said so explicitly.

They did not. This leaves SEP pool administrators operating under a framework designed for a structurally different market, and it leaves open a litigation vector that implementer-side counsel will not hesitate to exploit. Licensors and pool administrators should document the optional, FRAND-constrained nature of their pool structures clearly and early in any enforcement or dispute context.

The no-challenge clause treatment adds a practical compliance issue. Paragraph 297 states that no-challenge clauses in pool licenses, including termination-upon-challenge clauses, are likely to fall within Article 101(1). Pool administrators should review their standard license terms before May 1.

What This Means in Practice

For SEP licensors facing LNG-style collective negotiation, paragraph 322 is the most immediately useful development in the document. The Huawei v ZTE framework remains intact as a constraint on holdout strategies. The ACCC Apple Pay determination provides the best available analytical parallel for framing the downstream competitive harm argument when implementer coalitions seek to formalize collective non-engagement.

For pool administrators, the analytical gap on SEP pools is the most significant structural problem. Document the FRAND-constrained, optional nature of your pool architecture proactively. Review no-challenge clauses before May 1.

For implementer-side counsel, the absence of a safe harbor is the governing fact. Every LNG needs structural care from inception. The buyer cartel boundary is real and the ACCC case shows that competition authorities will look through efficiency arguments to the downstream harm.

For those watching the broader EU policy trajectory, the Commission declined to take a structural position on FRAND rate-setting, SEP valuation, or the licensor-implementer balance. The contested questions from the SEP Regulation debate are not resolved here. They are deferred to courts, arbitrators, and the next policy cycle.

That is not a retreat. It is a deferral. And in the standards world, deferrals have a way of becoming the permanent state of affairs.

Standards at Risk covers SEP licensing, FRAND policy, and the economics of technology standardization. If you found this piece useful, share it with someone navigating the new compliance landscape. The May 1 deadline is closer than it looks.

TL;DR

Three things the SEP licensing community needs to know right now:

• Guoling No. 835 is not an improvised reaction to DS611 or US tariff pressure. It is the capstone of a six-year legislative build — the sixth instrument in a deliberate sequence dating to 2020. This was planned, not reactive.

• The headline features — malicious entity list, asset freezes, trade bans — are real but not the most dangerous provision. Article 8(4) is.

• Article 8(4) prohibits Chinese organizations from sharing data or personal information with entities placed on the malicious entity list. In a cross-border licensing dispute, that prohibition maps directly onto the discovery and evidence-sharing processes that make enforcement possible. A licensor can win in London and find itself unable to build the evidentiary record needed to enforce or pressure settlement anywhere else. No court ruling required. No WTO challenge available. It is self-executing.

The Provision Nobody Is Discussing

Most commentary on Guoling No. 835 has focused on the headline features: the malicious entity list, asset freezes, trade bans, visa denials. Those are real and the SEP licensing community should be paying attention to them. But there is a provision buried in Article 8 that is, in practical terms, more threatening to foreign patent holders than any of those tools.

Article 8(4) authorizes prohibition or restriction on Chinese organizations and individuals from providing data or personal information to entities placed on the malicious entity list, or engaging in transactions or cooperation with them. That restriction applies not only to the listed entity itself but, per the final paragraph of Article 8, to any organization actually controlled by or established with the participation of a listed entity.

Read that again in the context of SEP licensing enforcement.

A patent licensor, its outside counsel, or a licensing pool administrator pursues enforcement of a UK global FRAND determination against a Chinese implementer. China formally identifies that determination as constituting unjustified extraterritorial jurisdiction under Article 6. The enforcing parties are placed on the malicious entity list. Under Article 8(4), Chinese parties — including subsidiaries, joint ventures, or any Chinese entity in a cooperative relationship with the licensor — are now prohibited from sharing data or personal information with those listed entities.

In a cross-border licensing dispute, that prohibition maps almost directly onto the discovery and evidence-sharing processes that make enforcement possible. The implementer’s Chinese affiliates cannot cooperate with document requests. Chinese entities in the licensor’s supply chain or licensing network cannot share transaction data. The evidentiary foundation of enforcement collapses — not because a court ruled against the licensor, but because a regulatory designation made the necessary information flows illegal.

This is not a theoretical risk. It is a structural mechanism for making foreign FRAND determinations practically unenforceable at the evidentiary level, without ever requiring China to challenge the underlying legal merits of the determination itself. The licensor wins in London. It cannot enforce in Shenzhen. And now it cannot build the evidentiary record it needs to pressure settlement anywhere else either.

The ASI Debate Is Now Academic

Whether Chinese anti-suit injunctions were a legitimate defensive response to UK extraterritorial overreach or an aggressive jurisdictional power grab was, for a time, a genuinely contested question. It no longer matters.

The ASI was a visible, case-by-case judicial tool. It generated a WTO dispute. It produced a ruling. It was withdrawn. That cycle is complete.

What Article 8(4) represents is a different order of instrument entirely. It does not appear in any court proceeding. It will not generate a WTO dispute. It operates upstream of litigation, enforcement, and rate-setting — in the information flows that make all of those processes possible in the first place. The debate about whether the original ASIs were justified or defensive is a debate about a tool that no longer exists, conducted in the shadow of a replacement that most of the commentary has not yet caught up to.

The question worth asking now is not whether China was right or wrong to issue ASIs. It is what the regulatory architecture that replaced them actually does — and whether the enforcement strategies currently being built in London, Munich, and Washington adequately account for it.

This Was Six Years in the Making

This is not an improvised response to DS611 or to US tariff pressure. The full legislative lineage makes that clear:

• 2020: Malicious Entity List rules

• 2021: MOFCOM blocking rules targeting secondary sanctions

• 2021: Anti-Foreign Sanctions Law

• 2023: Foreign Relations Law

• March 2025: State Council Decree No. 801, Regulations on Resolving Foreign-Related IP Disputes

• April 2026: State Council Decree Guoling No. 835, Regulations on Countering Unjustified Foreign Extraterritorial Jurisdiction

Each instrument filled a specific gap. Guoling No. 835 is the capstone: higher legal authority than any of its predecessors, consolidating identification, blocking, countermeasures, prohibition orders, and private litigation rights into a single unified framework. China’s Ministry of Justice was explicit in its official Q&A: the purpose was to fill structural gaps left by earlier instruments, specifically to address more complex and hybrid forms of external pressure that fell outside the scope of the prior toolkit.

The data transfer restriction in Article 8(4) is precisely that kind of hybrid tool — designed for exactly the scenario where a foreign court wins on the merits but cannot enforce in practice.

The Bottom Line

For patent holders building enforcement strategy around global FRAND determinations, Article 8(4) is the provision to understand. Not as a hypothetical. As a present operational reality.

Asset freezes and trade bans are visible, reversible, and negotiable. A prohibition on data sharing embedded in domestic Chinese law is none of those things. It is self-executing, durable, and it recruits every Chinese entity in the ecosystem as an unwilling participant in blocking enforcement.

The licensor wins in London. It cannot enforce in Shenzhen. And now it may not be able to build the evidentiary record it needs to pressure settlement anywhere else either.

Standards at Risk covers the intersection of patent licensing, global standards policy, and the strategic dynamics shaping the future of innovation markets. Part 1 of this analysis is available here.

The prevailing narrative around WTO dispute DS611 is that the EU challenged China’s use of anti-suit injunctions in SEP licensing disputes, the arbitrators ruled against China in July 2025, and China complied by withdrawing its ASI policy. That narrative is incomplete in ways that matter significantly for patent holders.

China’s withdrawal was real but carefully constructed. The Supreme People’s Court issued no prominent public notice, ASI doctrine has not been repudiated, and the clearest evidence of withdrawal came from China’s own statements at the WTO rather than any transparent domestic legal action. This is recalibration, not compliance.

More importantly, while attention was focused on DS611, China enacted two significant regulatory instruments that collectively replace the ASI with something more durable and harder to challenge.

The shift is from a visible, case-by-case judicial tool that was vulnerable to WTO challenge, to a systemic regulatory architecture that is prospective, administrative, and far harder to contest through dispute settlement. China did not retreat from the contest over who sets global FRAND rates. It upgraded its position.

For SEP licensors the practical consequences are direct: the enforceability gap between winning a global FRAND determination and collecting on it is widening; holdout behavior now has regulatory cover; and forum selection strategy must account for the real possibility that a favorable UK High Court ruling is unenforceable in the markets that matter most. A follow-on WTO dispute, DS632, has been filed by the EU challenging China’s authority to set binding global SEP rates without both parties’ consent, but that case will take years and faces the same structural compliance ambiguity that DS611 produced.

The deeper issue is jurisdictional fragmentation. The UK, the UPC, the US, and China are each deploying tools to assert or protect their own authority over global licensing outcomes, with no treaty framework to resolve the conflicts between them. The fight in SEP licensing is no longer primarily about what FRAND means. It is about who gets to decide.

A WTO Dispute Is Not a Court Judgment

DS611 was a formal WTO dispute, meaning a state-to-state complaint, not a private lawsuit. The EU filed against China alleging that Chinese anti-suit injunction practice violated TRIPS obligations around effective enforcement and access to judicial remedies. The WTO case page is at link.

It is worth being precise about what TRIPS actually requires in this area, because the obligations are less absolute than they are often described. The full WTO overview of TRIPS enforcement provisions is at link. The core provision is Article 41, which requires that WTO members make available enforcement procedures that permit effective action against infringement, with remedies that are expeditious and constitute a deterrent to further infringement. Article 41 also requires that these procedures be fair and equitable, and not unnecessarily complicated, costly, or subject to unreasonable delays. Critically, Article 41 is a minimum standards obligation. Members must meet the floor, but the agreement explicitly does not create an obligation to put in place a judicial system for IP enforcement distinct from enforcement of law in general.

The second obligation at issue in DS611 came from Article 63, which requires transparency: laws, regulations, and final judicial decisions of general application pertaining to IP must be published or otherwise made available. It was under Article 63 that the arbitrators found China had failed to publish or make accessible decisions, including Xiaomi v. InterDigital, that articulated principles guiding future cases. The full TRIPS text, including Article 63, is at link.

The EU’s argument was that China’s ASI practice, by blocking SEP holders from pursuing litigation abroad and imposing coercive financial penalties for doing so, effectively denied rights holders access to the enforcement procedures they were entitled to under Article 41. That argument ultimately prevailed on appeal, though not without real legal tension, as discussed below.

The process runs through several stages: consultations between the parties, establishment of a panel, issuance of a decision, and nominally an appellate stage. I say nominally because the WTO Appellate Body has been effectively dysfunctional since 2019, when the United States blocked the appointment of new members. That dysfunction hasn’t stopped dispute resolution entirely. DS611 proceeded under an Article 25 alternative arbitration arrangement, with the arbitrators issuing their award in July 2025, but it has changed the character of outcomes.

There are no damages. There is no direct enforcement mechanism. The losing party is expected to bring its measures into conformity; if it doesn’t, retaliation may be authorized. In practice, WTO outcomes are often political settlements wearing legal clothing. Governments adjust behavior without admitting wrongdoing, the complaining party declares satisfaction, and the case moves toward closure. That process creates exactly the kind of ambiguous compliance space China has now exploited.

One important nuance worth noting: the WTO panel in early 2025 initially rejected most of the EU’s substantive claims. It was the arbitral appeal, decided in July 2025, that reversed those findings and concluded that Chinese ASI practice was inconsistent with the TRIPS Agreement. Even then, the arbitrators’ reasoning on what constitutes a “final judicial decision of general application” that must be published under Article 63 has been legitimately criticized as stretching a vague TRIPS obligation beyond its institutional context. China does not operate a formal common law precedent system, and a first-instance provincial decision, even an influential one, does not carry binding effect across the Chinese judicial system. That is a real tension in the ruling, and acknowledging it honestly does not undermine the broader argument. It sharpens it: even a ruling with debatable reasoning produced only ambiguous compliance.

What DS611 Was Actually Contesting

The official framing of DS611 was about anti-suit injunctions, those orders issued by Chinese courts, most notoriously the Wuhan Intermediate Court in Xiaomi v. InterDigital, that blocked SEP holders from pursuing litigation abroad, backed by daily fines reaching RMB 1 million for non-compliance. The arbitrators concluded that Chinese ASI practice altered the process of FRAND negotiation in a fundamental way, diminishing SEP holders’ ability to conclude licensing agreements and impacting their negotiating position when filing suits outside China.

But the real dispute beneath the legal pleadings was simpler and more fundamental: who gets to set global FRAND rates?

When a UK court asserts jurisdiction to determine a global FRAND license, as it has done repeatedly, that is a claim of sovereign authority over royalty outcomes that affect Chinese manufacturers, Chinese supply chains, and Chinese industrial policy. From Beijing’s perspective, this is not neutral adjudication. It is extraterritorial reach into a domain China considers its own economic interest. The ASI was the blunt instrument China used to push back. It worked, for a while. But it was also visible, controversial, and ultimately untenable under WTO pressure.

The “Retreat” That Wasn’t

Here is what the record actually shows about China’s response.

The Supreme People’s Court reportedly issued a notice in September 2025 withdrawing the ASI policy and communicated this to the WTO. The EU published its acknowledgment of China’s withdrawal at link. On its face, that looks like compliance.

Look harder. As of early April 2026, the SPC’s notice was not discoverable through ordinary searches of the SPC website or the SPC IP Tribunal site, a pointed irony given that the dispute itself turned partly on judicial transparency obligations under TRIPS Article 63. The clearest evidence of withdrawal came not from a prominent SPC publication, but from China’s own statements at the WTO (link) and the EU’s reporting of those statements. Meanwhile, SPC and SPC IP Tribunal materials have continued to describe and even celebrate the development of ASI practice. The doctrine has not been dismantled. The cases have not been repudiated. The Wuhan government’s summary of Xiaomi v. InterDigital, proudly labeling it the “world’s first cross-border anti-suit injunction,” remains online at link.

China appears to have taken the position that there never was a written ASI policy to begin with, which conveniently makes withdrawal of a nonexistent policy consistent with maintaining the underlying doctrine. Whether an anonymized publication of a possibly non-permanent link constitutes meaningful compliance with TRIPS Article 63 transparency obligations is a question the arbitrators did not fully anticipate, and the SPC has not rushed to resolve it. Mark Cohen’s detailed analysis of this compliance gap is at link.

As of early April 2026, Mark Cohen, one of the most closely watched observers of Chinese IP practice, reported no awareness of new Chinese SEP anti-suit injunctions since the July 2025 ruling, though given the transparency failures at the heart of DS611 itself, absence of observed activity is not the same as confirmed absence.

This is not compliance. It is recalibration. China has done exactly enough to satisfy the formal WTO compliance process while preserving its ability to revisit these tools if circumstances change. And the reason it can afford to make that trade is that it has already built something more durable to replace them.

Two Regulations, One Strategy

While DS611 attracted most of the attention, the more consequential developments arrived in two legislative waves, the second of which landed just this month.

The first came in March 2025, when China’s State Council enacted Decree No. 801, the Regulations on Resolving Foreign-Related Intellectual Property Disputes. Analysis of that regulation is at link. Binding on both administrative agencies and courts, it constructed the initial framework: restrictions on cooperating with foreign evidence requests, authorization for countermeasures against countries using IP procedures to “contain or suppress China,” and prohibitions on Chinese entities assisting foreign governments in what the regulation characterizes as discriminatory proceedings. Articles 12 through 17 were the operational core, with trigger conditions deliberately broad enough to cover most global FRAND determinations by a foreign court.

The second wave arrived April 7, 2026: State Council Decree Guoling No. 835, the Regulations on Countering Unjustified Foreign Extraterritorial Jurisdiction. The full Chinese text is at link, with an English translation available at link. Signed by Premier Li Qiang and effective immediately upon publication, this is 20 articles of purpose-built blocking architecture. It is not an amendment or extension of the 2025 regulation. It is a distinct and more operationally aggressive instrument, and it goes materially further in three respects that matter directly for SEP licensing.

First, it establishes a malicious entity list, a formal registry of foreign organizations and individuals deemed to participate in implementing extraterritorial jurisdiction measures against China. The trigger for listing is not criminal conduct. It is participation in enforcing a foreign measure that China has formally identified as constituting unjustified extraterritorial jurisdiction. A patent licensor, its counsel, or a licensing pool administrator enforcing a UK global FRAND determination against a Chinese implementer is not obviously outside the scope of that provision.

Second, it authorizes a graduated countermeasure toolkit: asset freezes, trade bans, visa denials, and blacklisting, deployed against foreign entities that harm Chinese interests through extraterritorial reach. These are not aspirational. They are named instruments in a regulation that is already in force.

Third, and most consequentially, it creates a private right of action, allowing Chinese citizens and organizations affected by foreign extraterritorial measures to bring lawsuits against those enforcing them in Chinese courts, with government authorities directed to provide guidance and support for such actions. A Chinese implementer subject to a UK global FRAND determination could potentially invoke this provision to sue the licensor’s enforcement counsel in China. That is a structural deterrent to enforcement that operates independently of any government action. It is self-executing litigation risk embedded in domestic law.

Taken together, the two regulations represent a coherent strategic architecture: the 2025 instrument built the policy framework, and the 2026 instrument gave it enforcement teeth. The question of whether a given foreign proceeding constitutes “unjustified extraterritorial jurisdiction” sits with the State Council legal affairs department, which may conduct investigations and issue public announcements. Once a measure is formally identified, Chinese organizations are prohibited from implementing or assisting in its enforcement, and may only proceed upon State Council approval.

One analyst characterized the 2025 regulation as China preparing for “IP lawfare.” The April 2026 regulation is not preparation. It is the opening position.

The Structural Shift: From Judicial to Regulatory Tools

The strategic logic deserves to be made explicit, because it represents a genuine evolution in how China engages this space.

Anti-suit injunctions were a judicial instrument. They were issued case-by-case, required individual enforcement, and were highly visible. That visibility made them politically costly, as they handed the EU a concrete, documentable grievance suitable for WTO dispute settlement under TRIPS Article 41.

The new regulatory framework is different in kind. It is systemic rather than episodic, administrative rather than judicial, and prospective rather than reactive. A Chinese company facing a UK global FRAND determination no longer needs a court to issue an ASI. It can invoke the regulatory framework to decline compliance, require State Council approval before taking any action under a foreign judgment, and potentially trigger countermeasures against the opposing licensor, including placement on the malicious entity list.

The shift is from blocking individual lawsuits to blocking the enforceability of entire categories of foreign decisions. That is a more durable position and, critically, one that is far harder to challenge through WTO dispute settlement, which is designed to address specific identifiable measures, not pervasive regulatory frameworks operating under broad sovereign authority claims.

The Complicating Factor: Nobody Has Clean Hands

ASIs are not a Chinese invention, and the jurisdictional picture has grown genuinely complicated in ways that deserve honest acknowledgment.

US and UK courts have used anti-suit injunctions for decades. The UPC has now developed anti-anti-suit injunctions, orders designed to prevent parties from seeking or enforcing foreign ASIs that would interfere with UPC proceedings, issued in disputes including Huawei v. Netgear and Broadcom v. Realtek in late 2024. In early 2025, the UPC’s Mannheim Local Division issued an ex parte anti-interference measure in InterDigital v. Amazon. The UK High Court responded with its own counter-order, restraining enforcement of those continental measures insofar as they affected UK proceedings. Multiple jurisdictions are now deploying overlapping tools to protect their own authority over rate-setting and enforcement.

The tool itself, in other words, is not the problem. What distinguished the Chinese ASIs in DS611 was a specific combination of factors: thin jurisdictional nexus, with Chinese courts asserting global rate-setting authority over portfolios with limited Chinese connection; aggressive running financial penalties designed to coerce rather than compensate; and proceedings issued without adequate notice to the SEP holder. The problem was the strategic purpose, which was systematically suppressing licensor access to legitimate foreign forums in violation of the Article 41 minimum standards that TRIPS guarantees. The new regulatory framework pursues the same objective through machinery that is harder to see and harder to challenge.

It is also worth noting that the USTR’s position in DS611, arguing against expansive transparency obligations under TRIPS Article 63, was a departure from over thirty years of bipartisan US policy encouraging China toward greater judicial openness. The arbitrators rejected that approach, which was the right outcome. But it is a reminder that the Western position on these disputes has not always been internally coherent, and that critics who point to Western jurisdictional overreach are not entirely wrong. The difference is that Western courts’ claims to global rate-setting authority, whatever their faults, are grounded in parties voluntarily submitting to jurisdiction as an alternative to injunction. China’s blocking framework is coercive by design. It operates without consent and against parties who have no meaningful connection to the Chinese regulatory process.

DS632: The Fight Continues

China’s partial compliance with DS611 has not ended the WTO track. In February 2026, the EU requested a panel in DS632, a follow-on challenge to China’s claimed authority for its courts to set binding worldwide SEP licensing terms, including royalty rates covering non-Chinese patents, without both parties’ consent. The DS632 case page is at link. The WTO established the panel on March 19, 2026. A substantial roster of third parties has reserved rights: Australia, Brazil, Canada, India, Japan, Korea, the UK, the US, and others.

DS632 goes to the core question that DS611 danced around. It is not about the procedural tool, the ASI, but about the substantive claim: does a Chinese court have the authority to set global rates for a patent portfolio that includes rights registered in other jurisdictions, without consent from both parties? The EU’s answer is no. China’s implicit answer, embedded in its prior practice and now given regulatory architecture, is yes, and the April 2026 regulation has effectively made that answer self-enforcing.

DS632 will play out over years. But its establishment confirms that the WTO track remains live and that the jurisdictional competition is not resolving. It is escalating. And the new blocking regulations will complicate DS632’s practical impact in exactly the same way they complicated DS611’s: China can adjust surface behavior while embedding structural resistance in domestic law.

What This Means for SEP Licensing

For patent holders, the practical implications are not abstract.

The enforceability gap is widening. A global FRAND determination favorable to a licensor is worth less in a world where the implementer can invoke a State Council regulation to block compliance, require government approval for any action under a foreign judgment, or expose the licensor’s enforcement counsel to placement on a malicious entity list. The legal win and the commercial outcome are increasingly decoupled.

Holdout behavior now has regulatory cover. The new framework does not merely give implementers leverage in individual negotiations. It creates systemic friction that shifts the balance in every negotiation involving a Chinese party. Holdout was already a structural problem in SEP licensing. It just got a regulatory backstop with enforcement teeth.

Forum selection has never mattered more, or been more constrained. The traditional playbook for patent holders, choosing favorable forums, obtaining injunctions or global FRAND rates, and using that leverage to drive settlements, depends on enforcement. As enforcement becomes jurisdiction-dependent and cross-border enforceability degrades, the value of favorable rulings declines accordingly. A UK High Court global FRAND determination is an increasingly powerful piece of paper in London and an increasingly uncertain one in Shenzhen.

It is also worth noting, against the persistent claim that SEP licensing regulation is addressing a real innovation harm, that the empirical literature does not support that premise. Research in the competition policy literature has found no credible evidence that SEP licensing practices have suppressed innovation or harmed the development of standards. The regulatory interventions now reshaping this space are driven by distributional concerns and industrial policy, not evidence of market failure. That matters because it clarifies what is actually at stake: sovereign control over the economics of innovation markets, not consumer protection.

The Fragmentation of Global SEP Enforcement

After two decades in this space, I have watched the SEP licensing debate move through several phases, from early FRAND rate-setting disputes, through the hold-up vs. holdout wars, through pool governance debates, to the current moment where the underlying question is no longer doctrinal. It is jurisdictional.

The UK courts assert authority to set global FRAND rates. The UPC is developing anti-interference tools to protect its own proceedings. China has withdrawn its ASIs under WTO pressure while constructing a two-layer regulatory blocking framework. The US position remains in flux. There is no single forum that controls global licensing outcomes. There is no treaty that allocates jurisdiction over global FRAND rate-setting. There is no enforcement mechanism that operates across all relevant markets.

What exists instead is a fragmented landscape of competing authorities, each with its own tools and each motivated to expand its jurisdictional reach. The WTO can adjudicate whether specific practices violate TRIPS. It cannot answer the deeper question: which court, in which country, has the right to determine what a global standard is worth?

The Question Is No Longer What FRAND Means

For most of the last two decades, the fight was about doctrine: FRAND rates, comparables, injunctions, smallest salable patent-practicing units. Those questions remain important. But they have been overtaken by something more fundamental.

The real fight is about who decides. Which courts have the authority to set global rates? Whose judgments are enforceable across jurisdictions? Which regulatory framework gets to define what counts as discriminatory IP practice?

China did not retreat from that contest when it pulled back its ASIs. It moved to higher ground, from a position legally vulnerable under WTO frameworks to one structurally more durable and far harder to challenge. The 2025 regulation built the policy architecture. The April 2026 regulation gave it enforcement teeth. Neither is a placeholder. Together, they represent a long-term strategic position on who controls the economics of global standards.

The fragmentation of global SEP enforcement is not a risk on the horizon. It is the current reality. And the question facing every participant in this ecosystem, licensors, implementers, courts, and policymakers alike, is not whether to engage with that fragmentation, but how.

For patent holders who have built their businesses on the premise that global standards create global licensing rights, that question has never been more urgent.

Standards at Risk covers the intersection of patent licensing, global standards policy, and the strategic dynamics shaping the future of innovation markets.

Parts 1 and 2 built the foundation. Part 1 established the multi-agent analytical framework and produced a directional prediction: confidential global license, not a public judgment. Part 2 corrected a documented error — the stayed Delaware companion case — added live PACER docket data for all four US proceedings, and surfaced the Western District of Texas cloud infrastructure case as a first-order variable nobody was discussing publicly.

Part 3 adds three analytical dimensions that the prior seven-agent system was structurally unable to model:

The validity challenge layer — whether pending PTAB proceedings could invalidate patents central to InterDigital’s enforcement campaign before key litigation milestones, and what that means for settlement economics.

The regulatory and competition law dimension — whether InterDigital’s patent acquisition strategy and multi-forum enforcement posture create Article 102 TFEU exposure, and whether Amazon’s abuse-of-dominance claim at the UK High Court is a throwaway pleading or a meaningful settlement chip.

The standards essentiality boundary — the threshold test under the ITU Common Patent Policy for ITU-T/ITU-R/ISO/IEC, the encoding/decoding fault line, and why this unresolved legal question may be the most important variable the UK trial will be forced to address.

To model these dimensions, three new agents were added to the swarm: a PTAB/Validity Agent, a Regulatory/Competition Agent, and a Standards Essentiality Agent. The system now runs ten agents across five rounds, producing approximately 57 API calls per run at a total cost of roughly $3.50 at current Claude Sonnet 4.6 pricing.

Before presenting the results, this article makes explicit corrections to both the legal framing and specific analytical claims from Parts 1 and 2. Three factual errors in the prior analysis are corrected here directly.

Three Factual Corrections from Parts 1 and 2

Correction 1: UPC contempt sanctions are not procedurally accruing.

Parts 1 and 2 described UPC contempt sanctions as “procedurally ripe and accruing.” That was imprecise. The Mannheim local division’s 27 February 2026 breach finding makes sanctions theoretically available, but a formal per diem rate requires a separate procedural step that has not been initiated. The merits appeal hearing scheduled for May 2026 is the primary reason enforcement has not progressed. The sanctions are a real threat — but describing them as currently accruing misstated the procedural posture.

Correction 2: The Western District of Texas case cannot produce remedies before September 2026.

Part 2 identified the Texas case as a secondary driver with the suggestion that it could produce injunctive relief or a damages award on a timeline preceding the UK trial. That was wrong. The Texas complaint was filed in February 2026. Amazon’s response is due in early May. No court produces injunctive relief or a damages award within four months of an answer to a complaint. The Texas case is a genuine long-term commercial threat to Amazon’s cloud infrastructure — but its timeline is 2027 or 2028 at the earliest, not 2026.

Correction 3: A UK video codec rate determination would not affect InterDigital’s wireless portfolio.

Parts 1 and 2 suggested that a below-market UK rate determination could “cascade as a de facto ceiling across its entire licensing portfolio.” That overstated the scope. InterDigital’s 4G/5G/6G wireless patents — its primary revenue base — are governed by different standards, different patent families, and different licensing programs. A UK determination on video codec patents under the ITU Common Patent Policy would have no direct effect on InterDigital’s wireless licensing program. The benchmarking risk is real but limited to the video codec licensing program specifically.

Two Legal Framing Corrections

Correction 1: This dispute is not straightforwardly a “FRAND dispute.”

Parts 1 and 2 used “FRAND dispute” and “FRAND negotiations” as shorthand for the entire proceeding. That framing accepts Amazon’s characterization of the legal question before it has been resolved.

InterDigital contests whether FRAND obligations attach to a material portion of its asserted portfolio. Its position is that the relevant video codec standards — H.265/HEVC and related specifications — specify decoder behavior only. Encoder implementations are deliberately left unspecified by the standard. Patents covering encoder-side implementations are therefore, in InterDigital’s view, not subject to any licensing commitment under the ITU Common Patent Policy, because those patents are not “required to implement” the Recommendation.

Amazon’s position is the opposite: that commercial reality makes these patents functionally indistinguishable from patents that are required, and that the UK court should set a rate covering the full portfolio.

The UK proceedings are simultaneously a rate-determination proceeding for patents where the FRAND commitment is undisputed, and a conventional patent enforcement dispute for patents where the FRAND obligation is actively contested. Calling the whole thing a “FRAND dispute” collapses a contested legal question into an assumed answer. The more accurate framing throughout is “the dispute over whether and to what extent licensing obligations under the ITU Common Patent Policy attach to InterDigital’s asserted portfolio.”

Correction 2: The applicable policy is the ITU Common Patent Policy — not ETSI’s IPR Policy.

Parts 1 and 2 referenced ETSI’s IPR Policy when describing the FRAND obligation framework. That was imprecise. H.265/HEVC is developed under ITU-T and ISO/IEC JTC 1 — it is governed by the Common Patent Policy for ITU-T/ITU-R/ISO/IEC, a joint policy of ITU, ISO, and IEC. ETSI’s IPR Policy is a separate instrument governing ETSI standards, primarily in the mobile telecommunications space.

This distinction matters for the UK proceedings. The extensive body of UK FRAND jurisprudence — Unwired Planet v Huawei, Optis v Apple, InterDigital v Lenovo — developed in the context of ETSI-governed 3GPP telecommunications standards. Whether that jurisprudence applies equally to ITU Common Patent Policy-governed video codec standards is a question that has not been definitively resolved by any UK court. Judge Meade is likely navigating relatively uncharted waters on this specific point.

The ITU Common Patent Policy: What It Actually Says

Having now reviewed the full Guidelines for Implementation of the Common Patent Policy (December 2022 revision), three points are analytically important.

The essentiality threshold is “would be required to implement.”

The policy defines a “Patent” subject to licensing declaration obligations as those claims “solely to the extent that any such claims are essential to the implementation of a Recommendation | Deliverable. Essential patents are patents that would be required to implement a specific Recommendation | Deliverable.”

The operative test is “would be required to implement” — not “commercially important,” not “practically necessary,” not “widely adopted.” Required. For HEVC encoder-side patents, where the standard specifies only decoder conformance and leaves encoder implementation free, the argument is straightforward: those patents are not required to implement H.265. An encoder can use any method to produce a compliant bitstream. Therefore encoder-side patents fall outside the policy definition of “Patent” entirely — no declaration obligation arose, and no FRAND commitment attaches.

The Organizations do not evaluate or certify essentiality.

The Guidelines are explicit: “The Organizations should not be involved in evaluating patent relevance or essentiality with regards to Recommendations | Deliverables.” The ITU patent database is described as “simply raising a flag to alert users that they may wish to contact the entities who have communicated Declaration Forms.” InterDigital’s declarations are statements of its own belief that its patents may be required — not admissions that they are. Amazon must affirmatively establish at the UK trial that each asserted patent meets the “would be required” threshold.

The assignment encumbrance is not legally absolute.

The policy states that licensing declarations “shall be interpreted as encumbrances that bind all successors-in-interest as to the transferred Patents” — but immediately qualifies this: “Recognizing that this interpretation may not apply in all jurisdictions.” This qualification is directly relevant to the Edgio acquisition. Whether any prior licensing declarations made by Edgio on the content delivery patents bind DRNC Holdings as successor-in-interest is a question of US patent law, not ITU policy, and bankruptcy sales can present complications for the enforcement of such encumbrances that the policy itself acknowledges it cannot resolve.

The PTAB Layer: What the Validity Challenges Actually Mean

The PTAB dimension entered the analysis because of a specific fact that surfaced in research: Dolby Laboratories filed an IPR petition against InterDigital’s US Patent No. 9,185,268 in February 2026. Both Dolby and InterDigital are licensors in the Access Advance HEVC patent pool. They are normally on the same side of the licensor table.

That fracture in the licensor coalition is strategically significant — but the PTAB/Validity Agent correctly identified that its direct relevance to the Amazon dispute is attenuated. Patent 9,185,268 is being asserted against Walt Disney Company, not Amazon. The same is true of the Unified Patents ex parte reexamination of InterDigital US Patent No. 10,805,610, granted by the Central Reexamination Unit in August 2025. That patent was also asserted against Disney.

This precision matters. The Dolby IPR and Unified Patents reexamination are not direct validity threats to the ten patents asserted against Amazon. They are indirect signals — evidence of portfolio fragility that Amazon’s counsel will use to pressure InterDigital’s negotiating position, and evidence that the licensor coalition is not monolithic.

The more important PTAB variable is what Amazon has not done. Amazon has a demonstrated willingness to file IPR petitions — it filed against Nokia in May 2024 (IPR2024-00847 and IPR2024-00848). It has not filed IPRs against InterDigital’s asserted patents in the current dispute. That is a deliberate strategic choice, and the PTAB/Validity Agent identified three likely reasons: estoppel risk in the parallel ITC and district court proceedings, the dramatically harder institution environment under USPTO Director John Squires (institution rates dropped from 68% in 2024 to below 40% by December 2025), and a deliberate decision to hold the PTAB option in reserve pending the May 2026 UPC ruling.

The PTAB environment under Director Squires actually favors InterDigital as patent owner. A Dolby IPR petition filed in February 2026 faces a discretionary denial rate of approximately 60%. Even if instituted, a final written decision is 12–18 months away — 2027 or 2028. That timeline does not threaten InterDigital’s 2026 enforcement campaign. But it creates a background signal that the portfolio is being tested, and that signal reaches every implementer InterDigital is currently negotiating with.

The Regulatory Flank: The Edgio Acquisition in EU Competition Law Context

Part 2 introduced the Edgio acquisition as a story about premeditated arsenal assembly. Part 3 runs it through an EU competition law lens, with the additional nuance that the ITU Common Patent Policy’s assignment encumbrance may or may not bind DRNC Holdings depending on jurisdiction.

The core facts: InterDigital acquired content delivery patents from Edgio Inc.’s bankruptcy through its subsidiary DRNC Holdings in January 2025 — seven months before Amazon filed in the UK in August 2025. These are non-SEP patents covering internet traffic management and routing. InterDigital has no organic presence in this technology area. They were deployed in the Western District of Texas in February 2026, targeting Amazon’s cloud infrastructure rather than its devices.

The EU competition law question is whether acquiring non-SEP patents against a specific counterparty during licensing discussions — and deploying them in litigation while simultaneously pursuing rate-setting proceedings for SEP patents — constitutes improper conduct under Article 102 TFEU. The Regulatory/Competition Agent took a committed position: the pre-dispute timing of the acquisition is legally significant.

The January 2025 acquisition predates the August 2025 Amazon UK filing. If Amazon’s counsel can demonstrate that InterDigital was assembling non-SEP enforcement capacity against Amazon specifically before the dispute formally commenced — using patents acquired from a bankrupt CDN company with no connection to video codec standards — it has a credible argument that this constitutes patent privateering deployed as a flanking mechanism. The EU Commission has flagged portfolio aggregation tactics in its SEP Regulation proposal. The 24 December 2025 EC notification means the Commission is already aware of the dispute.

Amazon’s abuse-of-dominance claim at the UK High Court adds a second dimension. This claim was Amazon’s opening move in August 2025. It has not been disposed of. If it survives, it could result in damages flowing to Amazon from InterDigital — which fundamentally changes the settlement economics. The Regulatory/Competition Agent identified this as Amazon’s most underappreciated settlement chip: Amazon can trade away the abuse-of-dominance claim in exchange for a rate concession or a broader release covering the Edgio/DRNC content delivery patents.

The pro-licensor framing remains valid — InterDigital is legally entitled to acquire and assert non-SEP patents, and building portfolio coverage against a specific implementer is legitimate strategy. Whether the Edgio acquisition creates Article 102 exposure is genuinely unresolved. But the regulatory flank is real, and it is the variable that could constrain InterDigital’s enforcement posture before any court rules on the merits.

The Standards Essentiality Boundary: The Most Underanalyzed Fault Line

This is the dimension generating the most analytically novel output in Part 3, and the one receiving least attention in public commentary on the dispute.

Under the ITU Common Patent Policy, the essentiality question reduces to a single test: would this patent be required to implement the Recommendation? Not is it commercially dominant, not is it widely adopted — would it be required.

For HEVC and related codec specifications, the standard specifies decoder behavior. What a compliant decoder must do when it receives encoded content is specified in detail. The encoder is deliberately left unspecified — any method that produces a compliant bitstream is permissible. This is the architectural design choice of every major video codec standard: decoder specification enables interoperability; encoder freedom enables innovation and competition.

The consequence for InterDigital’s portfolio is direct. Encoder-side patents cover methods implementers choose to use — methods that produce more efficient compression, better quality, lower bitrates. But the standard does not require any particular method. An implementer could use a different encoder approach and still produce a conforming bitstream that a standard-compliant decoder will correctly decode. Under the “would be required to implement” threshold, encoder-side patents are not required. They fall outside the policy definition of “Patent.” No FRAND commitment attaches.

InterDigital’s argument follows this logic precisely: the dispute over encoder-side patents is not a FRAND rate-setting matter. It is a conventional patent infringement dispute where InterDigital is entitled to negotiate on commercial terms.

Amazon’s counter-argument is that commercial reality collapses the distinction — no one building a commercial streaming device would use a significantly inferior encoder, making the dominant encoder-side patents functionally required even if technically optional. The UK court would be asked to determine whether “would be required” means technically necessary or commercially necessary, and that question has not been resolved under ITU Common Patent Policy jurisprudence.

There is a genuine technical nuance worth surfacing here. Modern encoders use internal decoding as a quality check within the rate-distortion optimization process — the encoder compresses a frame, internally decodes it to measure distortion, then adjusts encoding parameters accordingly. Whether this internal use of standard-specified decoder behavior brings encoder-side patents within the “would be required” threshold is a question no court has definitively answered. The Standards Essentiality Agent correctly flagged it as legally open.

The HDR patents are the most legally vulnerable subset of the portfolio on essentiality grounds. High dynamic range processing — metadata handling, tone mapping, dynamic range conversion — occurs on both sides of the pipeline. The clean encoder/decoder distinction that protects InterDigital’s encoding patents is harder to sustain for HDR patents where the standard specifies behavior on both sides. These are the most likely candidates for judicial scrutiny on the “would be required” question.

The strategic implication is significant: if a material portion of InterDigital’s asserted patents fall outside the “would be required” threshold, Amazon’s FRAND rate-setting jurisdiction argument weakens at the UK court. Judge Meade may be forced to address the essentiality question patent by patent, in the context of an ITU Common Patent Policy framework that UK courts have less experience interpreting than ETSI’s policy. InterDigital has structural reasons to prefer a confidential settlement over a public judgment that sets potentially unfavorable precedent on this question for its entire licensee portfolio.

The Revised House View

The ten-agent system produced the following structured prediction after five rounds, a coverage audit, and final refinement:

Prediction: The parties reach a confidential global license agreement in July or August 2026 — in the 60–90 days immediately preceding the UK High Court’s September trial. The deal covers the decoder-side technically essential patents in InterDigital’s HEVC/AVC/VVC portfolio across Amazon’s Fire TV, Echo, Kindle, and streaming device lines, structured as a lump-sum back-payment plus per-unit running royalty. The rate is constrained from below by Access Advance pool rate coherence — InterDigital cannot discount so deeply as to destabilize existing licensee relationships — and from above by PTAB attrition risk and the broader portfolio fragility signal. The ITC investigation is withdrawn by consent. The UPC AILI proceedings are dissolved. The encoder-side and content delivery patent claims are resolved through a separate commercial agreement or folded into a broader release as a condition of closing. Amazon’s abuse-of-dominance claim is traded away in exchange for a rate concession. No court issues a published rate. The settlement architecture is structured to satisfy UPC Rules of Procedure — a non-trivial constraint given the Mannheim local division’s 11 February 2026 ruling. The UPC breach finding from 27 February 2026 creates meaningful enforcement risk for Amazon, but contempt sanctions require a separate procedural step and have not been formally initiated pending the May 2026 merits appeal.

Alternative Outcome: The UK High Court proceeds to judgment in late 2026 or early 2027, issuing a rate on a structurally narrowed portfolio — encoder-side patents excluded or deferred pending separate essentiality determination, HDR patents subject to additional scrutiny, the Edgio/DRNC content delivery patents treated as outside the FRAND framework entirely. The published rate is technically binding but commercially incomplete. Amazon accepts under protest and appeals on portfolio scope. The ITC investigation is suspended but not withdrawn. The parties negotiate a supplemental commercial agreement on excluded patents within 12 months of judgment. The published rate becomes a floor that weakens InterDigital’s position in subsequent negotiations with Apple, Samsung, and Google. Probability: approximately 25–30%.

Key Driver: The UPC Court of Appeal ruling on AILI scope, expected May or June 2026. If the ruling narrows or dissolves the AILI, Amazon loses its primary European procedural shield and faces simultaneous ITC risk, Munich ASI enforcement, and UK trial exposure. Its walk-away number collapses and settlement in the July–August window becomes near-certain. If the AILI is upheld or expanded, Amazon retains sufficient fragmentation leverage to absorb UK trial risk and the alternative outcome becomes operative.

Secondary Driver: Whether InterDigital’s internal projection of the UK court’s likely rate on the decoder-side technically essential portfolio exceeds Amazon’s current settlement offer. This is private information, unobservable from the outside, and the one variable that could override the convergence logic even if the AILI is dissolved. If InterDigital believes Judge Meade will come in high on the decoder-side patents, it has rational incentive to proceed to judgment and use the published rate as a benchmark for its video codec licensing program — accepting short-term delay for long-term portfolio valuation. The Western District of Texas case is a genuine long-term commercial threat to Amazon’s cloud infrastructure, but on a 2027–2028 timeline — it cannot produce remedies before the September 2026 UK trial and should not be treated as a 2026 forcing mechanism.

Main Uncertainty: The settlement architecture must satisfy UPC Rules of Procedure, but the May 2026 UPC Court of Appeal ruling may itself redefine what those requirements are. A settlement valid under the pre-appeal interpretation may require renegotiation after the ruling issues. This procedural constraint — the most underappreciated variable in the dispute — could delay execution even after both parties have agreed in principle on rate.

Confidence: 68%.

What Changed Across Three Runs

The directional prediction has held across all three iterations: confidential global license, not a public judgment. What has changed is the mechanism, the timing, the legal precision, and the confidence.

Part 1 predicted Q4 2026, driven primarily by the €50M UPC penalty. Confidence: 52%.

Part 2 corrected the Delaware error, added the full US docket picture, surfaced Texas as a first-order variable, and moved the settlement window to late 2026 or early Q1 2027. Confidence: 54%.

Part 3 adds the validity challenge layer, the regulatory flank, the standards essentiality boundary, three factual corrections from Parts 1 and 2, and two corrections to the legal framing — replacing ETSI with ITU Common Patent Policy, and replacing “FRAND dispute” with a more precise characterization of the contested essentiality question. The settlement window sharpened to July–August 2026. Confidence: 68%.

The most important analytical improvement across all three runs is the identification of the “would be required to implement” threshold as the load-bearing legal question — not the technical/commercial essentiality binary, but a single policy test that InterDigital argues excludes its encoder-side patents from FRAND obligation scope entirely. That argument, if it prevails even partially at the UK trial, narrows the portfolio subject to judicial rate-setting and creates exactly the precedent risk that drives InterDigital toward a confidential settlement.

The Adversarial Critic’s dissent — that InterDigital’s private rate projection is the true unobservable determinant — remains the most important single insight across the entire series. It correctly identifies that the convergence narrative can be overridden by private information held by one party, and that this risk is structurally irreducible regardless of how much external data the system ingests.

A Note on the Methodology

Part 3 adds two new data sources alongside the three new analytical agents. The USPTO Open Data Portal PTAB API was queried live for all identified InterDigital proceedings and key asserted patents. The API returned no indexed results for the most recent filings — the Dolby IPR petition from February 2026 appears too recent to be fully ingested into the public database. Verified research data on the Dolby IPR and Unified Patents reexamination was sourced from IP Fray (10 February 2026) and Unified Patents (8 August 2025).

The legal framing corrections in this article are grounded in the primary source documents: the ITU Common Patent Policy Declaration Form and the Guidelines for Implementation of the Common Patent Policy for ITU-T/ITU-R/ISO/IEC (December 2022 revision), both reviewed directly.

The total cost of the Part 3 run — 57 API calls across ten agents, five rounds, coverage audit, and final refinement — was approximately $3.50 at current Claude Sonnet 4.6 pricing. Runtime was approximately seven minutes.

The full Python implementation — all three pre-processors, the ten-agent swarm architecture, and all enriched input builders — is available at github.com/theharlans/standards-at-risk.

Sources: JUVE Patent, “InterDigital vs Amazon — A chronology of the escalation,” 16 March 2026. CourtListener/PACER docket data retrieved April 2026. IP Fray, 10 February 2026. Unified Patents, 8 August 2025. Bloomberg Law and Mondaq for Texas and ITC proceedings. USPTO Open Data Portal PTAB API queried April 2026. Guidelines for Implementation of the Common Patent Policy for ITU-T/ITU-R/ISO/IEC (December 2022 revision) and Patent Statement and Licensing Declaration Form reviewed directly. All analysis is the author’s own. This is not legal advice.

After Part 1 published, a well-sourced critic posted publicly that my analysis had committed an “idiocy that can be demonstrated very easily.” He attached a court document showing the Delaware case I had flagged as a potential pressure lever was stayed — and had been stayed for years — as a companion to an ITC investigation. A stayed companion case carries no independent coercive weight. He was right.

I conceded the point publicly and immediately. But I also drew a distinction that matters: the system reasoned over the JUVE Patent chronology as its seed document. That article did not reference the Texas or Virginia cases, or the ITC investigation. That is a source document limitation, not an AI hallucination. The system can only reason over what it is given.

The correct response to a source document limitation is better inputs. So I rebuilt the input layer.

This is Part 2.

What Changed in the Methodology

Part 1 used a single source document — the JUVE Patent chronology — as the seed for the entire six-agent analysis. That article covered the European and UK proceedings in detail but mentioned only one US case, and mentioned it in passing.

For Part 2, I added a pre-processing step: a Python script that calls the CourtListener/PACER API to pull live docket data for all US proceedings, combines that with the original JUVE article, and writes an enriched input file that the swarm then analyzes. The six-agent architecture, the seven roles, the five rounds, the coverage audit, and the final refinement all ran unchanged. Only the inputs improved.

What the CourtListener pull revealed was that the US proceedings picture is not one case. It is four — and they are materially different from each other in status, patent family, and strategic significance.

The US Proceedings Map: What Part 1 Missed

Here is what the corrected fact ledger shows:

District of Delaware — 1:25-cv-01365 — STAYED Filed November 7, 2025. Five video codec patents. This is the case the critic correctly identified — a companion complaint to the ITC investigation, stayed pending ITC resolution. Carries no independent coercive weight. Should not have appeared in the pressure analysis at all.

ITC Investigation — 337-TA-3869 — ACTIVE Filed December 2025. The same five video codec patents as Delaware. This is the primary US enforcement lever — not Delaware. An ITC exclusion order blocks physical import of FireTV, Kindle, and Echo Show devices into the United States. That is commercial disruption at the supply chain level, not a financial penalty. However — and this is the correction that materially changes the timing prediction — a December 2025 ITC filing puts the ALJ initial determination on a standard 15–18 month schedule, meaning approximately March–June 2027. Not Q4 2026.

Eastern District of Virginia — 2:25-cv-00822 — ACTIVE Filed December 18, 2025, originally in the Alexandria Division, transferred to Norfolk. Four different video coding patents: US 8,149,338; 11,252,435; 12,143,606; 12,149,734. This is a separate patent family running on an independent timeline, assigned to District Judge Arenda L. Wright Allen, represented by McKool Smith PC — a specialist patent litigation firm known for aggressive trial posture. This case was entirely absent from Part 1.

Western District of Texas, Midland/Odessa Division — ACTIVE Filed February 2026 by InterDigital and subsidiary DRNC Holdings. Six content delivery patents: US 7,921,259; 10,116,565; 8,745,128; 9,015,416; 8,868,701; 8,583,769. This is not a video codec case. It targets internet traffic management and routing — Amazon’s cloud infrastructure and streaming delivery platform. It was entirely absent from Part 1, and it is the most strategically significant US filing in the entire dispute.

The Edgio Angle: Premeditated Arsenal Assembly

The Texas case contains a detail that the system correctly flagged in the Part 2 fact ledger and that deserves more attention than it has received anywhere else.

InterDigital acquired the content delivery patents through its subsidiary DRNC Holdings — purchased from Edgio Inc., a content delivery network company, in January 2025. Edgio went through bankruptcy. InterDigital, whose core business is video codec and wireless SEP licensing, has no organic presence in internet traffic management and routing technology. These are not patents InterDigital developed. They are patents InterDigital bought.

The acquisition date is January 2025. Amazon filed in the UK in August 2025.

InterDigital was assembling its litigation arsenal — including a completely different patent family specifically targeting Amazon’s cloud infrastructure — seven months before Amazon pulled the trigger on the UK proceedings. That is not reactive enforcement. That is premeditated multi-front preparation.

This reframes the escalation narrative in an important way. The conventional reading is that Amazon initiated by filing in the UK. But InterDigital had already positioned itself to attack Amazon’s most profitable business segment — AWS and Prime Video delivery — before the opening shot was fired. The Texas filing in February 2026 was not an escalation. It was the deployment of a weapon that had been loaded seven months earlier.

From a pro-licensor perspective, this is sophisticated portfolio strategy executed at the highest level. InterDigital identified Amazon’s commercial vulnerabilities across multiple technology layers, acquired the necessary patent coverage where its organic portfolio had gaps, and built a litigation architecture capable of threatening not just Amazon’s device business but its cloud infrastructure. That is what serious IP enforcement looks like.

What the Revised System Found

With the corrected, enriched input layer, the seven-agent swarm produced a materially different analysis in three specific areas.

On the ITC timeline: The system correctly identified that the December 2025 ITC filing produces an ALJ initial determination in approximately March–June 2027 — well past Q4 2026. The Part 1 prediction of a Q4 2026 settlement driven by ITC pressure was therefore built on a timing error. The ITC is a genuine lever, but it operates on a 2027 timeline, not a 2026 one.

On the Texas cloud infrastructure case: The system elevated this from a footnote to a first-order variable. A broad injunction or adverse ruling in Texas would not just affect Amazon’s device sales — it would threaten Amazon’s ability to serve third-party cloud and streaming customers through AWS. That is a completely different category of commercial disruption. The system’s US/ITC Agent, whose mandate is to evaluate whether US actions create meaningful leverage tied to physical products and supply disruption, correctly identified that the Texas case extends the threat surface from consumer hardware to Amazon’s most profitable business segment.

On the Mannheim breach finding: The system correctly weighted the 27 February 2026 UPC finding that Amazon was actually in breach of the ASI — not merely at risk of breach. Contempt sanctions and the €50M penalty are not hypothetical future risks. They are procedurally ripe now. This materially strengthens InterDigital’s near-term European leverage independent of the May 2026 appeal outcome.

The Revised House View

Prediction: Amazon and InterDigital reach a negotiated global patent license in late 2026 or early Q1 2027, before any binding judicial FRAND determination issues from the UK High Court. The settlement is not driven by any single forum but by the convergence of four simultaneous pressure vectors: UPC contempt sanctions that are procedurally ripe and accruing; the Texas cloud infrastructure case threatening AWS and Prime Video delivery at scale; the UK trial producing a rate anchor both parties can negotiate around; and InterDigital’s existential need to avoid a below-market public RAND determination that would cascade as a de facto ceiling across its entire licensing portfolio. The agreed rate lands above Amazon’s pre-litigation position but below InterDigital’s initial demand, structured as a multi-year global portfolio license with a retrospective lump sum. All active proceedings are withdrawn by consent upon execution.

Alternative Outcome: The UPC Court of Appeal narrows the AILI in May 2026, restoring Amazon’s fragmentation strategy. The ITC timeline slips past Q4 2026 without producing a near-term exclusion order threat. The Texas case is stayed or proceeds slowly. Amazon accepts the UK judicially determined rate under protest, challenges its extraterritorial scope in US courts, and the parties execute a license on judicial terms after a 6–12 month post-judgment enforcement phase. InterDigital suffers a below-demand rate that depresses its broader licensing program. Probability: approximately 30–35%.

Key Driver: The May 2026 UPC Court of Appeal ruling on AILI scope — specifically whether it upholds the AILI broadly and affirms that UK RAND damages proceedings improperly interfere with UPC jurisdiction. If it does, Amazon loses its primary delay forum and faces simultaneous coercive pressure from ripe UPC contempt sanctions, the Texas cloud infrastructure case, and the approaching UK trial rate anchor. Settlement becomes near-certain. If the Court of Appeal narrows the AILI, Amazon’s fragmentation strategy survives and the UK trial proceeds to a public judgment.

Secondary Driver: The trajectory of the Western District of Texas case. If it produces injunctive relief or a damages award on a timeline that precedes the UK trial, Amazon’s exposure extends from consumer hardware to its most profitable business segments and its negotiating posture collapses regardless of the ITC or UPC outcomes.

Main Uncertainty: Whether the Texas content delivery case is stayed, proceeds slowly, or accelerates. This is the variable that most determines whether the settlement pressure is overwhelming or merely significant. It is also the variable that received the least public attention before this analysis.

Confidence: 54% on late 2026 / early Q1 2027 global license. 35% on alternative fragmentation path. 11% on outcomes outside these two scenarios.

What Changed From Part 1 and What Didn’t

The directional prediction survived: a confidential global license remains the most probable outcome. The mechanism and timing got sharper.

What changed: the ITC is no longer the primary forcing mechanism on a 2026 timeline. The Texas cloud infrastructure case is now a first-order variable. The settlement window moved from Q4 2026 to late 2026 / early Q1 2027. The confidence number moved from 52% to 54% — a modest improvement reflecting a more complete fact base rather than a fundamentally different analysis.

What didn’t change: InterDigital’s benchmarking risk remains the most important structural incentive driving both parties toward a confidential rather than public resolution. The May 2026 UPC appeal remains the central hinge point. The February 2026 near-settlement remains the most important data point — parties who were in touching distance once, blocked only by a procedural technicality, will find their way back.

The most important correction is the one the critic forced: the Delaware case is irrelevant. The ITC is real but slow. The Virginia case is active and underreported. The Texas case is the sleeper that nobody is watching.

A Note on the Methodology

Part 2 adds one new capability to the system: live docket data retrieval via the CourtListener/PACER API. Before the swarm runs, a pre-processing script pulls current docket metadata for all identified US proceedings, combines it with the source article, and writes an enriched input file. The swarm then analyzes the enriched input unchanged.

The CourtListener API returned docket metadata for the Delaware and Virginia cases. Docket entries at the filing level require an elevated access tier. The ITC and Texas dockets were included via verified public sources rather than direct API retrieval.

The total cost of the Part 2 run — 37 API calls across seven agents, five rounds, coverage audit, and final refinement — was approximately $2.50 at current Claude Sonnet 4.6 pricing. Runtime was under five minutes.

The full Python implementation — including the CourtListener pre-processor, the swarm architecture, and the enriched input builder — is available at github.com/theharlans/standards-at-risk.

Sources: JUVE Patent, “InterDigital vs Amazon — A chronology of the escalation,” 16 March 2026. CourtListener/PACER docket data retrieved April 2026. Bloomberg Law, Mondaq, and verified public sources for Texas and ITC proceedings. All analysis is the author’s own. This is not legal advice.

In mid-March 2026, JUVE Patent published a detailed chronology of the escalating dispute between InterDigital and Amazon. It was the kind of article that rewards close reading: ten patents, five jurisdictions, a chain of anti-suit injunctions and counter-injunctions stretching from Mannheim to London to Luxembourg, a €50,000,000 penalty order, a near-settlement in February that fell apart over a procedural technicality, and a European Commission notification tucked in on Christmas Eve.

The case is genuinely complex. It sits at the intersection of standard-essential patent (SEP) law, FRAND licensing obligations, post-Brexit jurisdictional competition between the UK High Court and the newly established Unified Patent Court, and the unresolved question of which institution gets to set global royalty rates for technology that underpins the entire streaming industry.

I wanted to produce a rigorous prediction of where this heads. Not a summary of what happened, but a structured view of what happens next — and why.

This piece is about how I built that analysis, what tools I considered and rejected, and what the process revealed about the limits of AI-assisted legal reasoning.

The Tool I Decided Not to Use

My first instinct was MiroFish, an open-source AI swarm intelligence platform built on the OASIS multi-agent framework from CAMEL-AI. The pitch is compelling: upload a document, spawn thousands of AI agents with independent personalities and memories derived from the seed material, watch them interact across simulated social platforms, and synthesize the emergent behavior into a prediction report.

For certain problems — predicting public opinion formation, modeling how misinformation spreads, simulating market reactions to an earnings shock — this is a genuinely powerful approach. The “God’s-Eye View” controls that let you inject new variables mid-simulation are particularly interesting for stress-testing assumptions.

But I asked ChatGPT whether it was the right tool for this problem, and the answer was no. The core issue: MiroFish’s power comes from emergent behavior across thousands of agents operating in a socially dynamic environment. InterDigital v Amazon is not that kind of problem. It has a small, defined set of decision-makers — two companies, three courts, one competition authority — each with clear institutional roles, legal constraints, and documented positions. Spawning a thousand agents to simulate a dispute with six actual stakeholders would introduce noise, not signal. The complexity of the tool would overwhelm the complexity of the problem.

What the problem actually required was a leaner, structured approach: a small number of agents, each assigned a specific perspective, forced to take positions rather than describe possibilities, and subjected to iterative adversarial critique.

How I Built the Analysis Instead

I implemented a multi-agent reasoning system in Python, calling Claude Sonnet 4.6 via the Anthropic API.

The methodology proceeded in five stages:

Stage 1 — Fact Extraction. The JUVE article was converted into a comprehensive fact ledger: every procedural event, every jurisdiction, every enforcement mechanism, every date. Nothing interpreted yet — just a structured reality map of what actually happened.

Stage 2 — Multi-Agent Reasoning. Six analytical roles were instantiated: the Licensor (InterDigital’s perspective), the Implementer (Amazon’s perspective), the UK Court (Judge Meade’s institutional logic), the UPC (the Mannheim local division’s institutional logic), a Strategic Analyst (outcome-focused, incentive-aware), and an Adversarial Critic (assigned to challenge every claim). Each agent was required to take positions, not hedge. The system was designed to force commitment — because comfortable ambiguity is the enemy of useful prediction.

Stage 3 — Iterative Rounds. The agents progressed through four rounds: initial positions, escalation path analysis, conflict and adversarial critique, and convergence toward a synthesis. Later rounds incorporated the outputs of earlier ones, so the final reasoning was built on layered pressure-testing rather than a single pass.

Stage 4 — Coverage Audit. A dedicated audit step identified missed facts, underweighted facts, and potentially irrelevant facts. This is the step that surfaces what the synthesis left out, which is often as important as what it included.

Stage 5 — Final Refinement. The final output incorporated the fact ledger, adversarial reasoning, audit corrections, and an explicit assessment of latent external factors: industry structure, institutional incentives, the broader SEP regulatory environment.

The output was a structured prediction document: a refined house view, an alternative outcome, a key driver, and a residual uncertainty — all with explicit reasoning chains, not just conclusions.

The Stress Test

Once the structured analysis was complete, I ran it through Claude as an independent reviewer — a separate instance with no access to the intermediate reasoning, asked to evaluate the analysis on its merits, identify blind spots, and produce its own structured view.

This is the part of the process I found most valuable.

The reviewer agreed with the directional thesis — that a confidential global FRAND license is the most probable end state — and validated the core structural reasoning about InterDigital’s benchmarking risk. But it pushed back in three specific places, and those challenges materially improved the analysis.

Those refinements, and the full revised prediction, are in the paid section below.

The Full Analysis: Prediction, Stress Test, and Revised House View

What the process ultimately uncovered — after two rounds of AI reasoning, an adversarial stress test, and a methodology pivot — was this: the most likely outcome is not a courtroom verdict. It is a confidential global license, quietly executed sometime in the first half of 2027, after the courts have done enough work to anchor a number neither side will admit they agreed to. The dispute will not end with a winner. It will end with a press release that says very little, and a rate that nobody publishes. The more interesting question is not whether that happens, but what breaks it: one specific ruling in Luxembourg in May 2026, and one track in the United States that has received almost no public attention.

The Case in Brief

InterDigital holds patents on video compression and HDR technology — the standards that make streaming services like Prime Video function. It claims Amazon infringes these patents across FireTV, Kindle, and Prime Video. When license negotiations over its Video Codex portfolio failed, Amazon did something unusual: rather than waiting to be sued, it went to the UK High Court in August 2025 seeking a declaration of non-infringement and — crucially — a FRAND rate determination under the UK’s established Unwired Planet framework.

What followed was a procedural arms race across five jurisdictions in seven months.

InterDigital obtained anti-suit injunctions from both the Munich Regional Court and the UPC local division Mannheim within weeks of Amazon’s UK filing. The UPC’s version — technically an anti-interim-license injunction (AILI) — was designed to prevent UK proceedings from affecting UPC patent infringement cases. Both orders were made ex parte, meaning Amazon had no opportunity to pre-empt them.

The UK responded with an anti-anti-suit injunction in October 2025, prohibiting InterDigital from taking further blocking action in other courts. That injunction has been upheld twice.

By the end of 2025, the dispute had produced: a €50,000,000 UPC penalty order, a European Commission notification, simultaneous infringement proceedings in Delaware, Rio de Janeiro, Munich, and Mannheim, and a UK RAND trial scheduled for September 2026.

In February 2026, the parties came within reach of settling the entire injunction battle. They were blocked — specifically — by the UPC’s Rules of Procedure, which the Mannheim local division interpreted as requiring Amazon to formally withdraw part of its UK damages claim in a legally binding way before any de-escalation agreement could proceed. The UK judge received that demand with “serious reservations.”

As of the date of this publication, that procedural standoff remains unresolved. The UPC Court of Appeal will hear the broader dispute in May 2026.

The Engine Behind the Prediction

For readers who want to understand the machinery, here is the implementation in enough detail to be replicable.

The Model Choice

Every agent call in this system used Claude Sonnet 4.6 via the Anthropic API. The choice was deliberate: Sonnet 4.6 sits at the right point on the reasoning-depth versus cost curve for this kind of iterative multi-call workflow. A single pass of six agents across four rounds, plus the coverage audit and final synthesis, produces 27 API calls. At Sonnet 4.6 pricing, the full analysis — including the independent stress test — costs approximately $1.50 to $3.00 depending on output length. Running the same pipeline on Opus would be roughly 5x more expensive for a task where the marginal reasoning gain is not worth it: the value here comes from the structure of the system, not from any single call being maximally intelligent.

The Commitment Instruction

The single most important design decision was the commitment instruction embedded in every agent’s system prompt. It reads, in part:

CRITICAL INSTRUCTION: You must take firm positions. Do not describe possibilities.
Assert what [your principal] wants, why, and what its next move is. If you are
uncertain, commit to the most probable path and say so explicitly. Never hedge
with 'it depends' without immediately resolving the dependency.

This instruction exists because language models default to epistemic caution — they describe possibilities rather than commit to predictions. That caution is appropriate in many contexts and fatal in this one. A prediction framework that produces “it depends on several factors” is useless. The commitment instruction forces the model out of description and into forecast.

Agent Instantiation

Each of the six agents is defined by a role prompt that specifies perspective, institutional incentives, and the specific lens through which it must read the dispute. Here is the Adversarial Critic definition in full — the role whose outputs were most analytically valuable:

Agent(
    name="Adversarial_Critic",
    perspective="Adversarial critic — assigned to challenge the consensus",
    role_prompt="""You are an adversarial critic. Your sole function is to challenge
the emerging consensus. You are not trying to be balanced — you are trying to find
the specific assumptions that are doing the most structural work in the prediction
and stress-test them.

CRITICAL INSTRUCTION: Identify the two or three load-bearing assumptions in the
current synthesis and attack them directly. For each assumption, state: what the
consensus assumes, why that assumption might be wrong, and what the prediction
looks like if the assumption fails. Do not agree with the consensus unless you
genuinely cannot find a credible challenge. Disagreement is your default position."""
)

The phrase “disagreement is your default position” matters. Without it, the critic tends to produce a balanced assessment that validates the consensus with minor caveats. That is not useful. The system needs a role that is structurally adversarial, not just nominally so.

Stateful Reasoning Across Rounds

Each agent maintains its own conversation history across the four rounds. This means that by round three, when the Adversarial Critic challenges the €50M penalty assumption, it is operating on a synthesis that has already been built up across two prior rounds of reasoning — not starting fresh. The API call structure looks like this:

messages = conversation_history + [
    {"role": "user", "content": round_def["instruction"]}
]

response = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=1500,
    system=system_prompt,   # Agent identity + seed article
    messages=messages,      # Full prior conversation for this agent
)

The key architectural point: agents do not see each other’s raw outputs within a round. They share only the round instruction. This prevents early-round groupthink — if the Licensor agent produces a compelling argument in round one, it does not contaminate the Implementer’s round-one reasoning. Convergence has to be earned through the iterative rounds, not borrowed from another agent’s prior output.

The Coverage Audit as a Separate Call

The coverage audit is a completely fresh API call — no conversation history, no agent persona, no prior context. Its system prompt is simply: “You are a rigorous coverage auditor for structured analytical outputs.” And its instruction is explicitly critical: “Do not praise the analysis. Your job is to find its weaknesses.”

This separation matters. If the audit were run as a continuation of the agent reasoning, it would be biased toward validating the synthesis it had just produced. By making it a clean call with an adversarial mandate, you get a genuine gap analysis rather than a post-hoc rationalization of what the agents already found.

The Independent Stress Test

The stress test is structurally independent in a way that is worth making explicit. The reviewer Claude instance receives the source article and the final structured prediction. It does not receive the agent role definitions, any of the four rounds of intermediate reasoning, or the coverage audit. This is not just a prompt instruction (”pretend you haven’t seen the reasoning”) — it is a genuine information barrier. The reviewer genuinely has not seen how the prediction was produced. Its job is to evaluate the output, not reconstruct the process.

What This Cost

27 API calls. Approximately $2.10 at current Sonnet 4.6 pricing for the full pipeline including the stress test. Runtime was approximately four minutes end-to-end on a standard laptop, dominated by API latency rather than local compute.

What the System Flagged Against Itself

Before the independent stress test, the system ran its own coverage audit — a dedicated step designed to identify what the synthesis had missed or underweighted. It is worth showing this directly, because it is the part of the process most AI-assisted analysis quietly omits.

On the €50M penalty, the audit noted:

“The penalty ceiling is treated as the primary coercive mechanism, but the 22 December 2025 order also states Amazon was ‘possibly already in breach’ of the ASI at that point — meaning the penalty exposure may have already begun accruing before the May 2026 appeal, which would materially accelerate Amazon’s settlement calculus beyond what the synthesis assumes.”

On the US proceedings, it flagged a structural omission:

“The US Federal District Court in Delaware is present in the facts and absent from the prediction logic. It could serve as additional leverage or complicate a global license structure.”

And on the core settlement assumption, the Adversarial Critic — one of the six agent roles built into the system — registered a dissent that was never fully resolved:

“The consensus assumes both parties’ incentives point toward settlement. The critic argues those incentives are asserted rather than demonstrated, and that InterDigital’s portfolio strategy may affirmatively favor a public judgment over a quiet deal.”

That last point did not change the prediction. But it did lower the confidence from 57% to 52% in the final synthesis — and it is the reason the alternative outcome carries more weight than a straightforward reading of the coercion mechanics would suggest.

These are not failures of the process. They are the process working as designed: surfacing the assumptions that are doing the most structural work in the prediction, so they can be examined rather than buried.

The Original Prediction

The multi-agent system produced the following structured view:

Prediction: Amazon and InterDigital execute a negotiated global FRAND license in Q4 2026, in the window between the UK RAND trial and formal judgment entry. The primary coercive mechanism is the UPC’s €50,000,000 penalty — potentially accruing since December 2025 — combined with the European Commission notification and InterDigital’s demonstrated willingness to use procedural surprise simultaneously across five forums. InterDigital’s pure-play licensing model limits its tolerance for a public RAND judgment that benchmarks its Video Codex portfolio for Apple, Samsung, and Google negotiations, creating a mutual interest in confidential resolution. No clean public judgment is entered.

Alternative Outcome: The UPC Court of Appeal narrows the AILI in May 2026, removing the accruing penalty as a live coercive mechanism. Amazon accepts the UK RAND rate for UK-validated patents only, takes a partial license, and contests UPC and US portfolio scope in parallel proceedings through 2027–2028.

Key Driver: Whether the €50,000,000 penalty is confirmed as an accruing per-period liability before the May 2026 appeal — converting it from a theoretical ceiling into a live financial exposure Amazon cannot defer.

Residual Uncertainty: Whether InterDigital genuinely prefers a confidential settlement over a public RAND judgment that it believes will come in at a rate high enough to serve as a favorable benchmark for future licensees.

Confidence: 52%

The Stress Test: Where the Analysis Holds and Where It Doesn’t

Running the output through an independent Claude review produced three substantive challenges.

Challenge 1: The €50M Penalty May Not Be Coercive for Amazon

The original analysis treats the UPC penalty as the primary coercive mechanism. The independent review challenged this directly: Amazon’s annual revenue exceeds $600 billion. A €50 million penalty ceiling is not existentially threatening — it is, at most, a rounding error in Amazon’s litigation budget.

What does matter is not the nominal ceiling but the accrual mechanism and enforceability. If the penalty accrues per-period and Amazon cannot credibly argue it has complied, the exposure compounds. If enforcement is stayed pending the May appeal, the number is theoretical. The correct framing is:

The decisive question is not the headline penalty figure. It is whether UPC-linked exposure becomes live, compounding, and operationally difficult to defer — and whether it does so before or after the May 2026 Court of Appeal hearing.

Challenge 2: The Q4 2026 Settlement Window Is Too Specific

The predicted timeline depends on a conditional chain: the May appeal confirms the AILI, the penalty begins accruing credibly, the September trial produces a rate, and both parties transact before formal judgment enters. Any single link breaking produces a different outcome.

The independent review argued that Q1–Q2 2027 is more defensible as the settlement window — after the September trial creates a rate anchor and after the May appeal clarifies AILI scope, but before US Delaware proceedings ripen into an exclusion order threat.

The February 2026 near-settlement is the most important data point here. Parties who were “in touching distance” once, blocked only by a procedural technicality, will find their way back once that technicality is resolved. The May 2026 UPC appeal is likely to resolve it — one way or another.

Challenge 3: The US Track Is the Most Underweighted Variable

Neither the original analysis nor its coverage audit gave adequate weight to the Delaware proceedings. InterDigital filed in the US Federal District Court in November 2025. US patent litigation — particularly if it evolves toward an ITC complaint — carries the threat of import exclusion orders against FireTV and Kindle devices. That is not a financial penalty. It is a commercial disruption that directly affects Amazon’s device business in its largest market.

An ITC exclusion order would create acute pressure on Amazon that dwarfs any European penalty ceiling in practical terms. This is likely the strongest single lever in InterDigital’s arsenal that has not yet been activated. It is also the reason the settlement window may slide later into 2027 rather than closing in Q4 2026: both parties may be waiting to see how US proceedings develop before finalizing terms.

The Revised House View

Incorporating the stress-test findings, the refined prediction is as follows:

Prediction: A confidential global FRAND license is executed, but not in Q4 2026. The most likely settlement window is Q1–Q2 2027, after the September 2026 UK trial produces a judicially determined rate anchor and after the May 2026 UPC Court of Appeal ruling clarifies AILI scope and accrual mechanics. The February 2026 near-settlement establishes that a deal structure exists; the remaining obstacle is price, not architecture. US Delaware proceedings ripening in 2026–2027 add the commercial urgency — device disruption risk — that converts bilateral willingness into executed agreement.

Alternative Outcome: The UPC Court of Appeal materially narrows the AILI in May 2026. Amazon proceeds to a public UK RAND determination in September 2026. InterDigital, concluding that a public judgment at a UK-anchored rate is more useful than continued uncertainty, allows judgment to enter. The dispute fragments into jurisdiction-by-jurisdiction enforcement through 2027–2028. No single global license is ever executed. This path carries approximately 30–35% probability — higher than the original analysis implied.

Key Driver: The May 2026 UPC Court of Appeal ruling on AILI scope. Not the penalty amount — the penalty’s enforceability and accrual mechanism. A full confirmation of the AILI makes settlement nearly inevitable. A material narrowing extends the timeline by at least 18 months and makes the alternative fragmentation path the more likely outcome.

Secondary Driver: The trajectory of US proceedings. If InterDigital moves toward ITC action on FireTV/Kindle, the commercial pressure calculus shifts decisively regardless of what the UPC does in May.

Main Uncertainty: Whether the UPC Court of Appeal treats the February 2026 impasse — where its own Rules of Procedure blocked a near-settlement — as evidence of institutional overreach, and pulls back from the AILI’s maximalist interpretation. This is an open legal question with no precedent. It is the single variable that most determines whether settlement occurs in 2026 or slides into 2027.

Confidence: 42% on Q4 2026 specifically. 68% on a global license executed before end of 2027. And 30% on fragmentation/attrition extending beyond 2027.

What This Process Revealed About AI-Assisted Legal Analysis

Three things stand out.

First, the multi-agent structure was genuinely useful for surfacing tension that single-pass reasoning misses. The Adversarial Critic role in particular forced the system to confront the assumption that the €50M penalty is coercive for Amazon — a claim that sounds compelling until you look at Amazon’s balance sheet.

Second, the coverage audit is the most underrated step. The most consequential gap in the original analysis — the US proceedings — was not in the main synthesis at all. It appeared in the fact ledger and was never carried forward. A structured audit step is the mechanism for catching that kind of structural omission before publication.

Third, running a separate independent review against the output is worth doing — not because it dramatically changes the conclusion, but because it forces you to distinguish between the parts of the analysis that are well-reasoned and the parts that are rhetorically compelling but empirically weak. In this case: the benchmarking risk reasoning was genuinely strong; the penalty-as-primary-lever framing was rhetorically clean but analytically underpowered. The stress test made that distinction visible.

The dispute itself remains open. The May 2026 UPC hearing is the next inflection point. If one wants to know what happens after that, one will need to run the analysis again.

Sources: JUVE Patent, “InterDigital vs Amazon — A chronology of the escalation,” 16 March 2026. All analysis is the author’s own. This is not legal advice.

The full Python implementation — agents, rounds, audit, synthesis, and stress test — is available at github.com/theharlans/standards-at-risk.