Proportionality Is the New Injunction Battlefield Before the UPC

From remedy to order

Few disputes over UPC injunctions still turn on whether they will issue. The argument has moved to the terms of the order.

The Court of Appeal has narrowed the first question with growing clarity. Meril v. Edwards reads Article 63(1) UPCA as more than a grant of power: once infringement is found, the Court shall prohibit its continuation unless there are special reasons for not doing so.¹ Philips v. Belkin makes recall, removal and destruction the rule under Article 64, to be refused only where they would be disproportionate.² In Fujifilm v. Kodak, decided in July 2026, the Court added that “patent rights are an accepted and lawful limitation of free competition” and that “an injunction naturally has a negative impact on the business of the infringer.”³ A defendant who resists an injunction because it would damage its business now argues against a firm line of authority, which reserves refusal for exceptional circumstances.

The same decisions show where the dispute has gone. An injunction can be legally available and badly designed. Which defendant must stop which act, involving which product, in which state, from which date, subject to which exception, verified by which evidence and enforced by which penalty? Those questions arise in every case, and the case law on proportionality answers more of them than its reputation suggests.

In these columns on 3 November, discussing the statistics behind the IPRED reform debate, I suggested that defendants should bring a proposed order alongside the defence.⁴ The present piece turns to what that order should contain. It is confined to final remedies before the UPC outside the FRAND setting, and reads the decisions as a guide to drafting relief.

A constrained remedial choice

Proportionality is a legal constraint, including on whether relief should be granted. Article 42(1) UPCA requires litigation to be dealt with in a manner proportionate to its importance and complexity, and Article 42(2) requires procedures and remedies to be used fairly and equitably and without distorting competition. Article 64(4) supplies a specific test for corrective measures, weighing the seriousness of the infringement against the remedies, the willingness of the infringer to convert the materials into a non-infringing state and the interests of third parties. Article 82(4) requires penalties to be proportionate to the importance of the order. R. 118.1 still refers to the Court’s discretion under Articles 63, 64, 67 and 80. The Court of Appeal’s statement that Article 64 leaves no discretion to refuse measures to which the patentee is entitled must therefore be taken together with its express exception for disproportionate measures. It does not exclude proportionality review altogether.⁵

These provisions place proportionality inside the order. They act on its scope, its timing and its means of enforcement far more often than on its existence.

The burden follows the same logic. In Philips v. Belkin, the Court of Appeal placed on the infringer the burden of pleading and proving disproportion and held that general assertions would not suffice.⁶ The Düsseldorf Local Division has applied the point twice. In M-A-S v. Altech, general objections to destruction on environmental grounds failed.⁷ In Cup&Cino v. Alpina, a small company’s warning of its own economic destruction failed for want of substantiation.⁸ Neither decision closed the argument. Both demanded facts connecting a burden to a feature of the order that could be changed.

Ordinary commercial harm does not, without more, establish disproportion. The stronger argument identifies the feature of the order that produces an excessive burden and proposes a term that avoids it while preserving effective patent protection.

Scope: product, act and patient

Scope is the first variable, and the decisions leave real margin.

In Meril v. Edwards, the Court of Appeal recognised that patient interests can justify an exception where the device is the only available treatment or offers a notable improvement over existing treatments. On the evidence, it excluded Meril’s XL transcatheter heart valves from the injunction and corrective measures where a physician confirms that the device is the only available treatment for an individual patient, while refusing an exception for intermediate sizes.⁹ The exception was tied to a product, a clinical condition and a documentary trigger. The Court treated the physician’s notification as sufficient and refused to make access depend on the patentee’s review.¹⁰ The carve-out works because it can be checked.

The Paris Local Division’s decision of 16 October 2026 in Align v. Angelalign shows the claimant’s side of the same exercise. The patent covered a system of aligners worn in sequence, only some of which carried the claimed torque ridge, so that enjoining those appliances alone could make the whole treatment ineffective.¹¹ The reasons record the limits the claimant itself accepted. The injunction would reach only treatments not yet begun or interrupted, ongoing treatments that would include no torque ridge in future would be left alone, and appliances already delivered to patients would escape corrective measures. Since the accused system was neither the only treatment nor a notable improvement over alternatives, the defendants established no further patient-based exception.¹² The limits the claimant had accepted did much of the work that a patient exception would otherwise have done.

Those limits appear in the reasons. The operative part orders recall, removal and destruction of the products it identifies, including products held by dental professionals.¹³ The gap matters once the order comes to be implemented. A limitation offered in the pleadings is safest when written into the operative wording requested, and its boundaries should not be left to an enforcement dispute.

Acts can be separated too. In M-A-S v. Altech, the Düsseldorf Local Division refused recall, removal and destruction for products that infringed only indirectly, while leaving open whether exceptional cases might justify them.¹⁴ The prohibition and the corrective measures do not travel together automatically.

Meril also shows how scope returns once enforcement comes into view. On appeal, under the heading of enforceability, Meril argued that an “and/or” in the description of the enjoined products made the order unclear. The Court of Appeal disagreed, because the parties agreed on the substantive scope and the patentee had itself accepted that the delivery device fell within the order only when combined with a valve having all the claimed features.¹⁵ The argument failed, but it shows where a loose phrase will be tested. An ambiguity is cheapest to remove before the order is made.

Time and the R. 118.8 clock

Time is the second variable. For the orders covered by R. 118.8, enforcement requires notification to the Court of the parts to be enforced, any necessary certified translation into the relevant official language, service of that notification and translation by the Registry, and satisfaction of any security requirement.¹⁶ In Philips v. Belkin, the Court of Appeal held that the request should specify a compliance period and that, if the decision sets none, the claimant must fix one in its notification.¹⁷ Its order in Kodak v. Fujifilm explains that the compliance period for merits orders runs only once the R. 118.8 requirements, including any security under R. 352, are satisfied. For provisional measures, the starting point is instead service of the order or the claimant’s notice, with any required security under R. 211.5 also relevant.¹⁸ Meril confirmed the distinction.¹⁹ The notification mechanism does not give a claimant freedom to impose an unreasonable period.

Within that frame, the divisions calibrate. In M-A-S v. Altech, Düsseldorf allowed 30 days from the notification for corrective measures and rejected an unsubstantiated request for four to six weeks.²⁰ In Cup&Cino, it extended the destruction period to three months, beyond what had been requested, because recalled stock first had to come back.²¹ In Align, Paris gave two months for information and 30 days for corrective measures, both counted from the notification.²²

These periods allow specified operations to be carried out. They authorize no continued infringing sales. The decisions examined here do not establish a general entitlement to a sales transition period, though neither do they settle every possible application for one. A defendant seeking such relief should specify its duration, volumes and customers, explain why no workable alternative exists, and offer reporting and security. Time needed to collect or destroy stock is a different request from permission to keep supplying the market.

Corrective measures, the rule and its edges

Corrective measures have their own proportionality text, and the Court of Appeal reads it strictly. In Philips v. Belkin, it held that recall, removal and destruction may be refused where they would be disproportionate, for instance for a minor infringement or where the infringer is willing and able to make the products non-infringing, although a serious infringement can still justify the measures.²³ In Fujifilm v. Kodak, it described Article 64 measures as the rule rather than the exception, declined the requested exclusion for printing plates kept as evidence, rejected a shelf-life exception to recall, and allowed a second recall.²⁴

Some limits concern the statutory scope of the remedy before proportionality is reached. Cup&Cino excluded advertising material from destruction because it was not a product within Article 64. M-A-S v. Altech treated corrective measures for indirect infringement as generally outside that provision, leaving exceptional cases open.²⁵ Publication is a separate remedy again: the Court of Appeal requires special circumstances before an infringer must publish its own declaration on its website, and Paris refused publication in Align because no legitimate interest had been shown and the other measures sufficed.²⁶

The willingness to convert, which Article 64(4) names expressly, deserves more use than it has had. A defendant able to remove an infringing firmware function or replace a component under audit should say so, and support the offer with technical evidence and a timetable the claimant can verify. The claimant, for its part, should test whether the proposal removes the infringing feature and whether its implementation can be checked. A recall from independent customers raises different questions from a stop on future shipments, and the request should identify the product population, its location, its owner and the channel through which it would return.

Penalties belong to the design

Recurring penalties are easily treated as an afterthought. They should be drafted with the prohibition.

The Court of Appeal’s order of 14 October 2025 in Kodak v. Fujifilm sets out the method. The threatened amount must deter non-compliance without exceeding reasonable limits.²⁷ The order should identify the triggering conduct, preferably by period, item or act, and may set a maximum that a later order can raise.²⁸ A penalty must first be ordered in an operative part, whether in the main decision or separately. Payment for an alleged breach then requires the Court’s assessment under R. 354.4 and does not follow automatically from the claimant’s allegation. The defendant bears the burden of proving full and timely compliance. Specifying the evidence in advance (an accountant’s confirmation, recall letters or independently certified destruction) reduces later disputes.²⁹ Confidentiality objections to that evidence should be raised during the merits proceedings.³⁰

The July 2026 merits decision in Fujifilm v. Kodak applied the method. Kodak faced a penalty of up to EUR 50 per square metre of printing plate or up to EUR 10,000 per day, at the patentee’s choice, and a separate daily penalty for corrective measures.³¹ The structure follows the product. A per-area trigger tracks volume for plates sold by surface, and a daily trigger tracks continuing conduct.

The Milan Central Division’s enforcement decision in Insulet v. EOFlow shows why the trigger must be precise. The Court of Appeal’s provisional order provided for up to EUR 250,000 for each individual violation. Milan read that wording as referring to the type of violation established, import or export, and not to each product. It considered a penalty for each individual transaction incompatible with the market and the parties’ size, and resolved the uncertainty in the infringer’s favour.³² That is a first-instance reading of a single order, and it lays down no general rule for penalty clauses. A claimant seeking a per-unit trigger should say so expressly and justify the resulting exposure.

Appeal, security and the cost of moving first

For infringement remedies, Article 74(1) UPCA provides that an appeal does not suspend enforcement unless the Court of Appeal orders otherwise on a motivated request. An appeal against a decision on revocation, by contrast, has automatic suspensive effect under Article 74(2). In Cup&Cino, the Court refused suspension, including an alternative request confined to recall and destruction, and rejected reliance on possible future events.³³ In Amycel, the partial irreversibility of recall letters and publication did not suffice to justify suspension.³⁴

A refusal to suspend does not end the risk analysis. In June 2026, the Court of Appeal set aside a Mannheim judgment in Kodak v. Fujifilm and ordered compensation for injury caused by enforcement.³⁵ That case concerned EP 3 551 174, a different patent from EP 3 476 616 in the July decision discussed above. Security under R. 118.8 and Article 82(2) protects a potential compensation claim, but its amount does not necessarily cap the eventual loss. Nor is it ordered automatically. Düsseldorf refused it in Cup&Cino after a case-specific balance, and Paris refused it in Align despite the claimant’s United States seat.³⁶

Both sides should model enforcement before judgment. The claimant should identify the states in which compliance will be sought, the translations it will need and the parts of the orders it will notify first, since R. 118.8 allows enforcement in parts. The defendant should prepare its compliance plan, its request for security and, if it has one, a specific ground for suspension. A remedy that looks proportionate in the judgment can carry a different risk once immediate enforcement and the appeal timetable meet national procedure.

The same method applies to orders already made. A patentee holding one should check that penalty triggers match the conduct, that compliance periods are stated and that the evidence of compliance is specified. A defendant should run the same checks, preserve evidence of compliance and raise genuine uncertainty before it becomes an alleged breach. Insulet rewards precise triggers. It does not make ambiguity a safe compliance strategy. Where enforcement proceeds pending appeal, both sides should assess the potential injury and the compensation that would follow if the decision were reversed.

The fear of conceding

The strongest objection comes from the defence side. A defendant that files a proposed order, with its carve-outs and periods, appears to concede the infringement it denies, and counsel are understandably reluctant to hand the panel a ready-made route to an injunction.

The objection weakens on inspection. A subsidiary request does not admit the main one, and auxiliary requests are a familiar feature of UPC pleadings. The front-loaded procedure leaves little choice in any event. Philips v. Belkin places the burden of pleading and proving disproportion on the infringer, and facts raised for the first time in a request for suspension or in an enforcement dispute come too late to shape the order. A defendant that stays silent on remedies does not protect its non-infringement case. It leaves the drafting of the operative part to the claimant.

The view from the claimant’s side

Most of the decisions discussed here were won by patentees. The lesson for them is precision. A broadly drafted request invites the enforcement disputes that the merits judgment seemed to have settled. A request drafted with its own limits, like the one in Align, removes the defendant’s best argument before it is made. Each product version identified, each act specified, each period stated and each item of evidence of compliance defined is one question fewer for the enforcement court.

The same discipline protects the claimant on appeal. A tailored order is harder to attack as disproportionate and easier to maintain if part of the case falls away.

The limits of the argument

Several questions remain open. The decisions do not settle when continued sales under controlled conditions might be justified, or on what conditions. Meril recognised the importance of a sole treatment or a notable clinical improvement, whereas Align found neither established. Whether similar reasoning could protect an infrastructure operator, or an industrial customer dependent on a single component, is a matter for evidence, and the medical cases cannot simply be extended to it.

In cases on standard-essential patents, the availability of an injunction depends first on the FRAND undertaking and on Union competition law, and a general appeal to proportionality cannot replace that analysis.³⁷ Proportionality should not become a catch-all for every objection. Each argument belongs somewhere: validity, infringement, FRAND, the provisional balance, the design of the final order, corrective measures or enforcement. It loses force when pleaded in the wrong place.

The injunction design matrix

Any request for final relief, or any response to one, can be built on a grid of ten fields: the right and the claims, the defendant or intermediary, the product and its version, the infringing feature, the prohibited act, the territory, the compliance date counted from the R. 118.8 notification, any transition or exception with its trigger, the verification and information to be supplied, and the corrective measures, penalties and security attached. Each field should rest on evidence and correspond to a sentence of the operative part requested.

The grid serves both sides. The claimant uses it to show that the order is precise enough to be enforced without further argument. The defendant uses it to locate the field in which an excess arises, and to propose a better term for that field alone, with the evidence that justifies it. The Court is then asked to choose between two drafts of the same order.

Proportionality rarely defeats an injunction before the UPC. It makes imprecision expensive. The operative part should therefore be drafted first, and the arguments on proportionality written to defend it.

¹ UPC Court of Appeal, 25 November 2025, UPC_CoA_464/2024 and connected appeals, Meril v. Edwards Lifesciences, paras 188 to 190. See Agreement on a Unified Patent Court (UPCA), Article 63.

² UPC Court of Appeal, 3 October 2025, UPC_CoA_534/2024, UPC_CoA_19/2025 and UPC_CoA_683/2024, Philips v. Belkin, paras 230 to 233 (the decision contains two paragraphs numbered 230). The patent concerned the Qi wireless charging standard. The decision is cited here for its holding on Article 64 UPCA.

³ UPC Court of Appeal, 13 July 2026, UPC_CoA_473/2025, UPC_CoA_474/2025, UPC_CoA_873/2025 and UPC_CoA_881/2026, Fujifilm v. Kodak, para. 385.

⁴ M. Dhenne, “94 Per Cent, 0.6 Per Cent: The Two Numbers Driving the IPRED Reform Debate”, Kluwer Patent Blog, 3 November 2026.

⁵ UPCA, Articles 42(1) and (2), 64(4) and 82(4). RoP, R. 118.1. Directive 2004/48/EC, Article 3(2). Philips v. Belkin, n. 2, para. 232. Meril v. Edwards, n. 1, paras 187 to 190.

⁶ Philips v. Belkin, n. 2, paras 236 and 237.

⁷ UPC Düsseldorf Local Division, 10 December 2025, UPC_CFI_316/2024 and UPC_CFI_547/2024, M-A-S v. Altech, para. 295.

⁸ UPC Düsseldorf Local Division, 18 March 2026, UPC_CFI_519/2024 and UPC_CFI_64/2025, Cup&Cino v. Alpina, para. 113.

⁹ Meril v. Edwards, n. 1, paras 191 to 197 and operative part, point X.

¹⁰ Ibid., paras 196 and 197 and operative part, point X.

¹¹ UPC Paris Local Division, 16 October 2026, UPC_CFI_684/2025 and UPC_CFI_2279/2025, Align Technology v. Angelalign, para. 126.

¹² Ibid., paras 127 and 128.

¹³ Ibid., operative part, points F to H.

¹⁴ M-A-S v. Altech, n. 7, headnote 1 and paras 314 and 315.

¹⁵ Meril v. Edwards, n. 1, para. 211.

¹⁶ RoP, R. 118.8.

¹⁷ Philips v. Belkin, n. 2, para. 240.

¹⁸ UPC Court of Appeal, 14 October 2025, UPC_CoA_699/2025, Kodak v. Fujifilm, paras 35, 36, 46 and 47, referring at para. 36 to Philips v. Belkin, para. 240.

¹⁹ Meril v. Edwards, n. 1, para. 212.

²⁰ M-A-S v. Altech, n. 7, para. 296.

²¹ Cup&Cino v. Alpina, n. 8, para. 118.

²² Align v. Angelalign, n. 11, operative part, points D and I.

²³ Philips v. Belkin, n. 2, para. 232.

²⁴ Fujifilm v. Kodak, n. 3, paras 394 to 397.

²⁵ Cup&Cino v. Alpina, n. 8, para. 117. M-A-S v. Altech, n. 7, paras 314 and 315.

²⁶ Fujifilm v. Kodak, n. 3, para. 399. Align v. Angelalign, n. 11, paras 144 and 145.

²⁷ Kodak v. Fujifilm, n. 18, para. 33.

²⁸ Ibid., para. 34.

²⁹ Ibid., paras 31 to 34 and 39 to 44. RoP, R. 354.4.

³⁰ Ibid., para. 45.

³¹ Fujifilm v. Kodak, n. 3, paras 383 and 398.

³² UPC Milan Central Division, 4 December 2025, UPC_CFI_1167/2025, Insulet v. EOFlow, points 4.3 to 4.7.

³³ UPC Court of Appeal, 24 March 2026, UPC_CoA_44/2026, Alpina v. Cup&Cino, paras 10 to 18, in particular para. 17. See UPCA, Article 74(1).

³⁴ UPC Court of Appeal, 16 January 2026, UPC_CoA_935/2025, Amycel, paras 11 to 14.

³⁵ UPC Court of Appeal, 2 June 2026, UPC_CoA_312/2025 and connected appeals, Kodak v. Fujifilm, operative part, points A and E.

³⁶ Cup&Cino v. Alpina, n. 8, paras 128 to 130. Align v. Angelalign, n. 11, paras 139 to 141.

³⁷ CJEU, 16 July 2015, C-170/13, EU:C:2015:477, Huawei v. ZTE, paras 60 to 71. On the relationship between the FRAND analysis and proportionality, see Dhenne, n. 4.