Download

The Afterlife of a FRAND Negotiation

Matthieu Dhenne (Dhenne Avocats)

October 22, 2026

A stone path curves through vermilion torii gates at Fushimi Inari Taisha in Kyoto, Japan.

A failed FRAND negotiation does not end when the parties leave the room. By then they have exchanged claim charts, discussed comparable licenses and moved some distance from their opening positions. One of them may have declined a proposal made by a neutral. Litigation resumes, and a question nobody asked at the outset becomes pressing: what, precisely, travels with them into the next forum?

The answer may matter as much as the royalty calculation. A technical analysis carried forward can save months of duplicated work, while a confidential concession carried into another forum can make the next mediation impossible. A refusal to settle may also resurface as evidence of unwillingness, although the proposal concerned a different portfolio or depended on a company that was never at the table.

The Patent Mediation and Arbitration Centre now offers an occasion to address the problem. Its Discussion Paper on guidelines for FRAND disputes in mediation, arbitration, expert determination and hybrid formats is open for comment until 26 October 2026, and a further public consultation on the draft guidelines has been announced.1 The paper presents itself as an outline, non-binding and descriptive, and it rightly treats the available procedures as modular. Modularity is precisely what makes the exercise difficult. A dispute that moves from one procedure to the next should not lose, on the way, the legal distinctions on which each procedure depends.

Settlement as evidence, settlement as sanctuary

Tokyo, Munich and the PMAC do not offer three versions of the same FRAND code. Their authority comes from different places, and so do the consequences of taking part.

The Tokyo District Court’s January 2026 guidance on SEP infringement litigation builds settlement into the proceedings themselves. In principle, the court recommends settlement at the first hearing and then schedules concentrated sessions for negotiation; it may ultimately present a proposal on global FRAND royalties, framed around the wishes of both parties rather than fixed unilaterally.2 This is a court-connected arrangement directed toward agreement. Its ambition should not be mistaken for a general power to impose a worldwide license once talks fail.

The companion guidance on SEP judicial mediation is more revealing still. It aims, in principle, at a settlement by the third session; a technical briefing at the second can move the final settlement session to the fourth. Where the case calls for written exchanges on non-infringement and invalidity, the parties are expected to file them promptly in any subsequent action. If mediation fails after a proposal has been presented, the proposal and the committee’s written opinion are attached to the record; the record also states when failure resulted from a party’s unwillingness to mediate. Either party may then produce that record in later proceedings, including provisional proceedings, on an abuse-of-rights defense.3 Settlement conduct is thus designed to shape the evidential setting in which an injunction will be assessed. The discipline this imposes is considerable, and so is the need to understand, before speaking, what may later be said about the discussion.

Munich’s Seventh Civil Chamber approaches the problem from the other end, through the adjudication of a FRAND defense. Its first-instance judgment of 5 February 2026 examines the implementer’s conduct in a set sequence, distinguishing outward from inner willingness, and treats payment of an undisputed minimum as appropriate in a case where the licensing scope and other terms had been agreed and only the royalty amount remained in dispute. The chamber’s guidelines of 13 August 2026 consolidate that judgment with three others into a general method.4 The question raised in these columns was the point at which predictability hardens into prescription. For the PMAC, the lesson is narrower. An expectation developed to decide an injunction cannot simply be transferred into a consensual procedure as the price of being regarded as cooperative.

The PMAC’s own Mediation Rules take the opposite view of what survives. Article 25 first imposes confidentiality on everyone involved, subject to limited exceptions that include the written agreement of the parties and the mediator. Unless the parties agree otherwise in writing or the law requires, Article 25(2) also prohibits disclosure of material prepared exclusively for the mediation and produced by another party or the mediator, views and suggestions of another party or the mediator, another party’s admissions, the mediator’s proposals, and indications that another party had or had not shown willingness to accept a settlement proposal. These matters cannot be introduced as evidence or made the subject of testimony in judicial, arbitral or similar proceedings.5 Read against the Tokyo model, the provision excludes almost exactly what the Japanese record is designed to carry. A mediator’s proposal should not quietly become the next tribunal’s benchmark. Nor should an unsuccessful mediation be converted into a report on which party deserves an injunction.

The contrast does not prove that one system works better. It reveals a choice about the function of settlement discussions, and the PMAC should make its own choice unmistakable.

The case for carrying conduct forward

The strongest objection to a sanctuary deserves to be stated in its best form. If nothing said in a mediation can ever be used, an implementer may treat the procedure as a costless delay, and a SEP holder may make a proposal it never expected to see tested. The Tokyo guidance names the risk openly, presenting the exclusion of both hold-up and hold-out as one of its purposes. Under the Article 102 TFEU framework in Huawei v. ZTE, both parties’ conduct matters to whether a dominant SEP holder may seek an injunction without abusing its position, and a procedure that conceals conduct may appear to work for whichever party has the least to show.6

The objection is serious, but the Rules already distinguish limited procedural information from settlement content. Where a request for mediation receives no response within the period set by the Centre, Article 8(4) allows the Centre, on a party’s request, to provide evidence of the attempt. Where the UPC has stayed litigation at the parties’ joint request to allow ADR, Article 6(2) requires the Centre to report whether the ADR proceedings ended in settlement, without revealing the settlement’s details. Neither provision authorizes a general assessment of a party’s conduct inside the mediation.

An optional evaluative record could not rest on Article 25(2) alone. A written agreement under that paragraph can relax its evidentiary restrictions, but disclosure must also comply with Article 25(1), including the mediator’s written agreement where that exception is relied upon. The receiving court retains control over admissibility and weight, and Article 25(5) separately restricts testimony by the mediator and Centre staff. The guidelines could help parties design such a record at the outset, with clear limits on its content and later use.7

Huawei v. ZTE supports a consensual determination of the royalty by an independent third party; it also preserves challenges to validity, essentiality and use. It does not itself authorize disclosure of mediation exchanges. A consensual route that later turns into a trap would betray the autonomy on which it rests.8

An obligation identified before a procedure

The choice of model begins before confidentiality. A negotiation can fail because the participants have never agreed on the legal question they are trying to resolve.

On a contractual account of FRAND, the starting point is the undertaking given in the standardization process. Under the ETSI framework, which French law governs, the IPR Policy, the licensing declaration and the implementer’s entitlement as beneficiary of a stipulation pour autrui provide the legal setting in which performance is owed, while the license that results remains a distinct agreement.9 Competition law keeps its independent corrective role, including in the assessment of exclusionary remedies. A contractual account does not make every forum competent, nor every proposed term FRAND.

Consider a pool negotiation. The administrator may have authority to offer a license without authority to submit every licensor to arbitration. A parent company may negotiate for products sold by its affiliates without holding a mandate to bind each of them to an award. These questions concern the design of the proposed solution and leave the case for consensual resolution intact.

In Tesla v. InterDigital, the UK Supreme Court allowed licensing claims concerning the Avanci platform to proceed. Its discussion of utility and fairness treated the platform as the essential party and left the participation of individual licensors to case management.10 That reasoning does not turn every pool member into a party to a future arbitration. The UPC Court of Appeal was equally careful in Vivo v. Sun Patent Trust. Its order of 16 March 2026 upheld the Paris Local Division’s decision to reject a preliminary objection and to defer the admissibility of a FRAND determination claim to the main proceedings, treating the determination as a step toward the injunction sought. It did not settle a general jurisdiction to fix global rates, and the later withdrawal of one of the Paris actions following settlement left the question open.11

Building on the Rules’ existing requirements on representation and scope, an early PMAC record could identify the undertaking at stake, the persons whose rights are engaged, the authority of those participating and the intended effect of any result. It should accommodate disagreement, because requiring the parties to resolve every jurisdictional issue before a meaningful discussion would defeat the purpose. Recording the issue is enough. It prevents an apparent settlement from dissolving when someone asks who can sign it.12

Evidence in transit

The same care should govern evidence. Suppose the parties jointly commission an essentiality sample during a mediation and later arbitrate. Is the sample binding, simply admissible or part of a protected settlement exercise? Does the expert’s conclusion concern the sampled patents only, or the portfolio as a whole? Silence is an expensive way to answer these questions.

The answers differ from one PMAC procedure to another. Under the Expert Determination Rules, a determination binds the parties unless they agree otherwise in writing, whereas an expert appointed by a tribunal informs a decision that remains the tribunal’s own.13 The guidelines should help parties choose deliberately between those effects and say, at the moment of choice, how any result will be used later. The Discussion Paper already addresses referrals to expert determination, consent awards and two Med-Arb models, and it mentions that a non-binding determination may serve as agreed evidence. The practical task is to turn these options into a clear protocol defining which evidence and conclusions may be reused, for what purpose and under which safeguards.

An optional transition agreement could separate reusable evidence from protected mediation exchanges. Existing technical documents would keep their ordinary legal status; passing them across a mediation table would not alone immunize them from production. Lawful reuse would still depend on privilege, contractual duties, third-party rights and any court order controlling onward disclosure. Comparable licenses are the obvious case, since their confidentiality clauses protect licensees who will never appear in the PMAC proceedings.14 Within those limits, the parties could agree on a common evidential record without treating every concession as a portable admission.

Willingness by design

A transition agreement would also help with the fashionable question of willingness to arbitrate. In Acer v. Nokia, the Court of Appeal of England and Wales considered a concrete arrangement combining an immediate, adjustable license with an arbitral route to final terms. The court held that Nokia’s offer met its RAND obligation and that the implementers’ refusal left their RAND claims with no real prospect of success. It therefore granted a stay, while rejecting Nokia’s separate challenge to jurisdiction. It expressly left open whether a mere offer of arbitration would suffice, and it reserved the position of an implementer with a substantiated objection to the proposed process.15 The judgment also proposed conditions for the stay that would preserve pleadings, disclosure, evidence and costs in a subsequent arbitration, subject to Nokia’s agreement and further submissions. The relevant inquiry concerns what the process will accomplish, the objections raised against it and the parties’ subsequent conduct. Counting offers to arbitrate reveals much less.

The PMAC could make credible offers easier to formulate by asking parties to specify scope, appointment, timetable, evidence, interim arrangements and the binding effect sought. It should preserve the distinction between a usable offer and an automatic safe harbor. Equally, an implementer’s refusal to accept a defective process should not become conclusive proof that it never intended to take a license.

Publication without a tariff

A further transition occurs when private proceedings feed public expectations. The PMAC and the European Patent Office have agreed to develop a database of case law on FRAND disputes, and the Discussion Paper itself hopes that anonymized publication will build a coherent body of case law.16 The ambition makes the question timely. An anonymized royalty figure can travel widely while shedding almost everything needed to understand it.

A useful record would preserve the question decided, the legal source of the obligation, the scope of the mandate and the material valuation assumptions, so far as confidentiality allows. Readers need to distinguish a settlement from an adjudicated outcome, an interim arrangement from final terms, and a court’s rejection of a defense from the determination of a license. Where anonymization prevents meaningful comparison, the limitation should be visible on the face of the entry.

The Rules already strike a balance. The Centre may publish awards and expert determinations in anonymized form unless a party objects in writing through the Centre’s case-management system within sixty days, a period that runs from notification of the award in arbitration and from the date of the determination in expert determination.17 A guideline should not enlarge those mechanisms, and an opt-out corpus should not be presented as a representative sample of FRAND disputes. Disputes whose publication happens to be possible offer a poor basis for a tariff; what such a corpus can offer is reasoning open to scrutiny.

The point extends beyond FRAND. Institutional unity does not by itself produce a shared legal method, as the debate on the distribution of UPC cases has shown.18 The PMAC can contribute by showing why particular procedures and outcomes were justified, while leaving courts, arbitrators and parties to assess their relevance elsewhere.

A legal reason to survive

The contribution I submitted to the consultation on 14 October turns these observations into proposed additions to the outline and its annexes, extending the procedural questions raised in June about the PMAC Rules.19

A failed negotiation need not leave the parties where they began. It may narrow the issues, settle technical facts, test valuation assumptions or identify a workable route to a license. The guidelines should make those gains available. They should also preserve the protections under which they were achieved. What survives a FRAND negotiation should have a legal reason to do so. Before deciding what a FRAND procedure should produce, the PMAC should decide what it allows the parties to leave behind.

Vermilion torii gates at Fushimi Inari Taisha, Kyoto, Japan. Photo: Jeremy Goldberg / Unsplash. Image adapted with AI-assisted retouching.