Dynamics Updates
When is a managing director an accomplice to patent infringement?
Philips v Belkin
UPC_CoA_534/2024, UPC_CoA_19/2025 and UPC_CoA_683/2024, decision of 3 October 2025[1]
This is an appeal against a finding of infringement by Belkin of Philips’ patent relating to wireless inductive power transfer, and against the dismissal of Belkin’s counterclaim for invalidity. In addition to injuncting several Belkin group companies, the first instance judgment ordered the managing directors of Belkin to “refrain from performing their services as managing director . . . in such a way that the patent-infringing acts are carried out” as accomplices to the infringement. It did not, however, injunct the managing directors more generally (i.e. if they were to start a new company to carry on infringement) or require any payment of damages by the directors.
Both Belkin and Philips appealed. Belkin requested that both decisions be set aside in their entirety, the infringement proceedings to be dismissed, and the patent to be revoked in Germany, Belgium, France, Italy, the Netherlands, Austria and Sweden. Belkin also expanded its arguments on first instance and claimed for the first time invalidity due to lack of inventive step. Philips requested that Belkin’s managing directors be ordered to cease and desist, provide further information and to pay damages, as well as requesting that Belkin pay the costs of the case and extend the recall following the finding of infringement.
The Court of Appeal dismissed Belkin’s request for dismissal of the counterclaim for revocation. In doing so, they did substantially consider Belkin’s new arguments, stating that they were “a matter of different understanding of disclosure passages in a document of the prior art on which Belkin had already based its argumentation in the first instance”, and that Phillips had sufficient opportunity to comment.
The Court of Appeal also dismissed Belkin’s appeal against the finding of infringement against the Belkin group companies. However, the Court did dismiss the order against the managing directors, finding that their activities were not sufficient to establish themselves as accomplices. The Court clarified that for a managing director to be an accomplice, they must either deliberately use the company to commit patent infringement, or if they are aware that the company is committing patent infringement, aware that the act of use is illegal, and takes no action to stop the infringement
[1] https://www.unifiedpatentcourt.org/en/node/149310
UKIPO Fee Increases from April 2026: What It Means for Your IP Strategy
Overview
The UK Intellectual Property Office (UKIPO) has announced its first major fee increase in years, effective 1 April 2026. This change will affect patents, trade marks, and designs, with fees rising by an average of 25%. While the adjustment reflects inflation and operational costs, it also supports investment in modernising services, including the introduction of the new OneIPO system. Despite these changes, UKIPO fees will remain among the most competitive globally, maintaining the UK’s reputation as a cost-effective choice for first filings.
When Was the Last Fee Increase?
The UKIPO has maintained stable fees for a remarkably long period. Patent fees were last increased in 2018, design fees in 2016, and trade mark fees have remained unchanged since 1998. This extended freeze means that even after the upcoming rise, UKIPO fees will still be relatively low compared to other major IP offices, maintaining the UK’s reputation as a budget-friendly jurisdiction for IP protection.
What’s Changing in April 2026?
From April 2026, the official fees for filing and maintaining IP rights in the UK will increase. For example, the online patent application fee will rise from £60 to £75, and the patent search fee will go up from £150 to £200. Renewal fees will see the most significant changes, with the cost of renewing a patent in its 20th year jumping from £610 to £810.
Why UK First Filing Still Makes Sense
Despite this, filing your patent application first in the UK remains a smart and competitive strategy. It provides an affordable route for claiming priority under the Paris Convention and offers access to the UKIPO’s relatively efficient examination process. Many businesses value this approach for its fairly speedy process and its cost advantages when compared to filing first in other jurisdictions such as the European Patent Office (EPO) or the United States Patent and Trademark Office.
However, filing first in the UK may not always be available, particularly if the invention was devised outside the UK, has foreign national inventors and/or applicants, or other local filing requirements apply. It is recommended to speak to an EIP patent professional for tailored advice.
Strategic Recommendations
With these changes on the horizon, businesses should act now to optimise their IP strategy:
File new applications before 31 March 2026 to lock in current fees
Renew early: up to three months in advance for patents, six months for trade marks and designs
Review IP budgets and update cost forecasts
Audit your portfolio to reassess non-core rights
What About the EPO?
The EPO typically reviews its fees annually, and while no official announcement has been made yet, an increase in 2026 can also be expected. EIP will continue to monitor developments and provide updates once details are confirmed.
Next Steps
We encourage clients to speak to EIP to plan filings and renewals ahead of the April 2026 deadline. Acting now can help mitigate the impact of these changes and ensure your IP strategy remains cost-effective.
Lucie Jones and Luke Galloway
Auxiliary claim requests inadmissible and indicative of likely invalidity in applications for provisional measures.
ONWARD Medical N.V. v. Niche Biomedical, Inc.,
UPC_CFI_693/2025 (Munich LD), decision of 17 October 2025[1]
In UPC_CFI_693/2025, the Munich Local Division of the Unified Patent Court refused a patentee’s request for preliminary injunctive relief against an alleged infringer, due to serious doubts about the patent’s validity. The case concerned European Patent EP 3 421 081 B1, titled “System zur Neuromodulierung” (“System for Neuromodulation”), which covers a medical device system delivering electrical stimulation to a patient’s nervous system in pre-programmed spatial and temporal patterns. The decision – issued on 17 October 2025 by the Munich local division (Panel 1, Presiding Judge Dr. Matthias Zigann) – addresses whether a patent proprietor can obtain interim injunctive relief based on amended (narrowed) patent claims when the original granted claims face validity challenges. In summary, the patent owner’s attempt to rely on auxiliary claim versions was rejected, and the application for an interim injunction was denied.
The claimant, ONWARD Medical N.V., is the registered proprietor of European Patent EP 3 421 081 B1, filed in 2017 and granted in 2020. The patent covers a neuromodulation system delivering pre-programmed electrical stimulation patterns to a patient’s nervous system. The defendant, Niche Biomedical, Inc. (doing business as “Aneuvo”), a US-based medtech company, markets the “ExaStim” neurostimulation system in Germany and France.
On 30 July 2025, Onward applied for a preliminary injunction under Article 62 UPCA, alleging that the ExaStim system infringed claim 1 of its patent. The application sought urgent injunctive relief, including a cease-and-desist order, product seizure, and disclosure of distribution details. The accused product included the ExaStim Stimulator, programmer, ReCure electrode pad, and cable.
Niche had already filed a protective letter on 25 July 2025 and, in response to Onward’s application, contested the injunction, arguing that claim 1 lacked novelty over prior art, particularly US 2016/0263376 A1 (“Yoo”). Niche asserted that these disclosures already taught the core features of the claimed system, including spatial and temporal stimulation patterns and variable parameters.
To counter the validity challenge, Onward submitted eight auxiliary claim versions on 29 August 2025, requesting that the court consider these narrowed claims as alternative grounds for granting the injunction. Niche opposed this, arguing that such amendments were procedurally improper in interim proceedings and further demonstrated the weakness of the granted patent.
The Munich Local Division denied the request for preliminary measures in its entirety. It found that Onward failed to meet the requirements for a UPC interim injunction, chiefly because the patent’s validity was not “sufficiently secure” in the court’s view. The auxiliary (alternative) claim requests were also refused as inadmissible and unjustified in this summary procedure.
1. Validity and Likelihood of Success
To grant a preliminary injunction, the UPC Agreement (Art. 62(4) UPCA) and the UPC’s Rules of Procedure require the court to “satisfy itself with a sufficient degree of certainty” that the patent is likely valid and infringed (in addition to considering urgency and balance of harm). The Court of Appeal has held that this standard means the judge must deem it “at least more likely than not” that the patent will withstand validity scrutiny (UPC_CoA_335/2023, order of 26 February 2024[2]). In this case, after a summary assessment, the Munich panel was not convinced that the patent-in-suit was more likely than not valid in its granted form.
Because the lack of likely validity was clear, the decision does not delve into a detailed infringement analysis or other PI factors. The failure to clear the validity hurdle was sufficient to reject the application.
2. Auxiliary Requests for Amended Claims
The most significant part of the decision addresses the patentee’s auxiliary requests (the attempt to salvage the injunction by narrowing the patent claims during the PI proceeding). The Munich Local Division took a strict stance on this issue:
Overall, UPC_CFI_693/2025 (Onward Medical v. Niche Biomedical) aligns with the principle that interim relief is an extraordinary measure – available only for patents that appear both infringed and likely to withstand validity challenges. The ruling also provides valuable clarity that auxiliary claim requests are generally futile in such proceedings, thereby encouraging patent owners to ensure their granted claims are litigation-ready if they intend to seek rapid enforcement. The Munich court’s approach in this case will likely serve as persuasive authority for other UPC divisions faced with similar attempts to introduce claim amendments at the PI stage, until the Court of Appeal weighs in definitively on this issue.
[1] https://www.unifiedpatentcourt.org/en/node/159629
[2] As reported on here: https://eip.com/global/latest/article/why_did_the_court_of_appeal_reverse_the_local_division_injunction_in_10x_genomics_vs_nanostring/
Court of Appeal of the UPC considers post-filed experimental data irrelevant to claim construction
STEROS GPA Innovative S.L. v OTEC Präzisionsfinish GmbH,
UPC_CoA_579/2025, order of 7 November 2025[1]
STEROS, the exclusive licensee of European Patent EP4249647B1 (“EP’647”), lodged an application for provisional measures against OTEC before the Hamburg Local Division (Court of First Instance, herein “CFI”) on 26 March 2025.
The patent relates to an electrolytic medium and claim 1 requires the presence (amongst other features) of solid particles, a conductive solution and a non-conductive fluid. The alleged infringer, OTEC, manufactures and sells electropolishing machines which include an electrolyte medium.
The dispute between the parties in relation to infringement centred on whether or not OTEC’s product included a non-conductive fluid as required by claim 1 of “EP’647”.
In the first instance proceedings, experimental data submitted by the parties showed that OTEC’s product comprised an emulsion-based fluid. The continuous phase of this emulsion was non-conductive, whilst the emulsion as a whole had a higher conductivity. STEROS argued that the presence of the non-conductive continuous phase meant that claim 1 was infringed. The CFI agreed and ordered the provisional measures[2]. The present order relates to appeal proceedings before the Court of Appeal of the UPC (CoA).
Claim Construction
Claim 1 of EP’647 did not specify a maximum value for the conductivity of the non-conductive fluid, but did specify a minimum conductivity of 10 μS/cm for the solid particles and conductive solution. In assessing infringement in the appeal proceedings, the CFI construed the “non-conductive fluid” feature of claim 1 as follows:
In the appeal proceedings, the CoA looked to the patent’s description for justification of the CFI’s approach and discussed the relevance of the experimental data.
The CoA disagreed with the CFI’s reasoning in (1). The description of EP’647 referred to the conductivity of emulsion-based non-conductive fluids as a whole and described in detail the significance of the non-conductivity of the fluid in the form present in the final electrolyte medium. In the case of OTEC’s product, this is the emulsified form. It is the conductivity of this emulsion that is described in the patent as crucial to achieving the effect of the invention. The fact that a component of this fluid is non-conductive is not relevant.
In relation to (2), the patent’s description was silent on the threshold value for a non-conductive fluid. STEROS submitted post-filed data relating to a fluid described as non-conductive in the patent, including measurements showing a conductivity greater than 10 μS/cm. They argued that this justified a relative assessment of conductivity by comparing the conductive solution and non-conductive fluids. By this reasoning, if the non-conductive fluid was significantly less conductive than the conductive solution, claim 1 was infringed. Crucially, the patent specification did not include any data for this example, nor was there a description of the proposed relative assessment of conductivity.
Additionally, the CoA stated that STEROS had measured conductivity prior to making the electrolyte medium, contrary to the teaching in the patent [as discussed above in relation to (1)]. OTEC also submitted data for the same example, following the protocol stipulated by the CoA and instead showing that the fluid had a negligible conductivity well below the value of 10 μS/cm.
The CoA was therefore not convinced by this line of argument and agreed with the CFI - the maximum value of 10 μS/cm value should be used to define non-conductivity. They further emphasised that experimental data not disclosed in the patent specification is “as a general rule, not relevant to the interpretation of the patent claims”. In the present case, this was particularly important because the way in which STEROS attempted to use the post-filed data was seen by the CoA to contradict the teaching in the patent specification.
Summary
In OTEC’s product, the conductivity of the emulsion-based fluid as whole was higher than 10 μS/cm and therefore was not “non-conductive”. The CoA considered the patent not infringed.
As the patent was found not infringed, the CoA set aside the impugned order. The application for a provisional injunction was rejected and STEROS ordered to pay the costs of the first instance and appeal proceedings.
[1] UPC_CoA_579/2025 (Court of Appeal of the UPC, Luxembourg), https://www.unifiedpatentcourt.org/en/node/159673
[2] UPC_CFI_281/2025 (Court of First Instance, Hamburg Local Division), https://www.unifiedpatentcourt.org/en/node/125946
Sustainable models: content licensing for generative AI training
It is fair to say that no technology in recent times has created so much excitement and controversy as generative AI, and in particular Large Language Models (LLMs). Generative AI companies and their backers have their sights on generating billions, if not trillions of dollars, directly or indirectly from their models. Models which, in the vast majority of cases, are trained on original works without permission from, or payment to, their authors. The works used for training are primarily obtained by scraping the open web, a uniquely vast repository of publicly accessible knowledge and information, constantly updated to reflect the latest developments in human knowledge and culture.
Even leaving aside the big questions of what is ethical and what is equitable, it could be argued that the generative AI industry must eventually move from being extractive to being sustainable. The AI companies need a continuing supply original works so that their models can continue to reflect an ever-changing world. And as we move to a world where more and more information is intermediated via LLMs (i.e. “answer engines”, rather than “search engines”), creators need a means of remuneration to enable them to continue creating and sharing their work, which is ultimately in everyone’s interest.
These issues haven’t escaped the scrutiny of the law. Thanks to a largely harmonised copyright system, creative works are by default protected by copyright wherever they are created and wherever they are subsequently used. But that harmony, and the copyright system itself, are now under threat, as different countries and regions, pursuing different policy goals, have started to diverge in their approach to generative AI. In the US – home ground of the most illustrious AI companies – battles are raging in the courts over what constitutes “Fair Use” of copyright works. In the EU, generative AI training has seemingly been grandfathered into a Text and Data Mining (TDM) exception that predates the technology, enabling training on copyright works provided the rightsholder has not explicitly opted out. Publishers and creatives argue that the opt-out is unworkable, while a recent case in Germany (GEMA v OpenAI) suggests that generative AI training can fall outside the TDM exception altogether. Under UK law there is less leeway for use of copyright works, though in a landmark ruling (Getty Images v Stability AI), it was found that importing a model into the UK trained on copyright works was not an infringement. The government is now considering its approach, aiming to balance its ambition to become an AI leader with the interests of its significant creative and media sectors.
Due to the scale of datasets needed for generative AI training (at least in its current guise), it may be unrealistic for AI companies to negotiate licences with each individual rightsholder. This cries out for a collective licensing scheme and a functioning data marketplace, perhaps analogous to that of the music industry. Another analogy is music and video streaming, where content creators are paid per stream. Something similar could potentially work for AI models, though it would require technology for attributing value either to models or outputs at scale. In either case, the volumes of data typically used in AI training could make such solutions costly and challenging to administer.
Collective licensing solutions do not fit well with the current dynamic and automated nature of web-scraping. An alternative, or additional, solution might involve integrating payment into web-scraping itself. During this process, a web-scraper requests a webpage via a URL, loads the HTML of the page (including JavaScript and all elements), and extracts and saves the data it finds. A list of URLs from which data has been extracted will be generated, and this information could be used to trigger payments to an account associated with the URL. This method would however rely on the companies sending out the scrapers to report honestly and transparently on what they have done, something which would be difficult to enforce or police.
Alternatively, web pages could require payment for web-scraper access. For example, software implemented on a web server could detect that a request for content is from a web-scraper, and in response, a payment element may first be returned which, if agreed to, will allow the scraper access to the URL. A solution along these lines has recently been implemented by US infrastructure company Cloudflare, which sits behind around a quarter of all websites. This solution would require less transparency and honesty from the AI companies, but would require operators to implement additional technology on their servers. To the extent that AI training (rather than web-scraping per se) is the relevant act, such a solution would also need to determine that the purpose of the web-scraping is to train an AI model. This may be critical as website owners would still wish to enable web scrapers for indexing by search engines to access their servers unimpeded. In the UK, a proposed amendment to the Data Access and Usage Bill, which would have legally required web-crawlers to be transparent about their purpose, was recently rejected. Innovative technical solutions around detecting and/or authenticating access by different web-crawlers may also be developed.
In our view, the most impactful approach to safeguarding creators’ IP rights is likely to involve a combination of technical and legal solutions. To the extent that courts and governments across the world decide that companies should require permission to use (and hence pay for) copyrighted content to train AI models, there will need to be scalable and workable technical infrastructure to support this. There is innovation in this space. But any technical solutions will work best if backed by strong copyright law. This will ensure that, even where attempts to circumvent the technology are made, there will remain the option of asserting authors’ IP rights in the courts.