The new EU AI Act – The 6 Questions You Want to Know

Aurilex • 1 April 2024

The EU AI Act will greatly impact the AI companies coming into the European market.

The European Union has long harbored ambitions of assuming leadership in the AI industry. As a component of its digital strategy, the EU aims to enact regulations on artificial intelligence (AI) to foster optimal conditions for the advancement and utilization of this pioneering technology.


In April 2021, the European Commission proposed the first EU regulatory framework for AI. AI systems with versatile applications are assessed and categorized based on the risks they pose to users. Varying levels of risk will correspond to differing degrees of regulation.


European Parliament members have voted in favor of the draft EU AI Act on 13th March 2024. The EU AI Act is poised to come into effect in the upcoming weeks, pending final procedural and linguistic checks. The implementation of this act will carry significant weight and impact in shaping the regulation of AI within the EU and globally.


  1. Who will be affected?


The AI Act has established precise definitions for various stakeholders in the AI landscape: providers, deployers, importers, distributors, and product manufacturers. This mandates accountability for all entities engaged in the development, deployment, importation, distribution, or manufacturing of AI models. Furthermore, the AI Act extends its jurisdiction to encompass providers and users of AI systems situated outside the EU, such as those in China, if the system's output is intended for use within the EU.


    2. What are the requirements of the Act?


The EU AI Act has a risk-based approach. It mandates that general-purpose AI models, including generative AI systems like large language models (LLMs) and foundation models, comply with a classification system organized into different tiers of systematic risk:


(Source: European Commission: Shaping Europe’s digital future)


Low-risk systems like spam filters or video games are subject to minimal requirements under the law, primarily entailing transparency obligations. This kind of system should inform users that the content is AI generated.


High-risk AI systems, such as autonomous vehicles, medical devices, and critical infrastructure (such as water, gas, and electric systems), necessitate developers and users to comply with supplementary regulatory obligations, including implementing risk management measures to ensure accuracy, robustness, and accountability framework incorporating human oversight.


AI systems that have adverse impacts on safety or fundamental rights will be classified as high-risk and will be categorized into two distinct groups:


1) AI systems that are used in products falling under the EU’s product safety legislation. This includes toys, aviation, cars, medical devices and lifts.


2) AI systems falling into specific areas that will have to be registered in an EU database:

  • Management and operation of critical infrastructure
  • Education and vocational training
  • Employment, worker management and access to self-employment
  • Access to and enjoyment of essential private services and public services and benefits
  • Law enforcement
  • Migration, asylum and border control management
  • Assistance in legal interpretation and application of the law.


Before being introduced to the market and throughout their lifecycle, all high-risk AI systems will undergo assessment. Individuals will retain the right to lodge complaints about AI systems with designated national authorities.


Prohibited AI systems, with few exceptions, include those presenting unacceptable risks, such as social scoring, facial recognition, emotion recognition, and remote biometric identification systems in public spaces.


Certain exceptions may be permitted for law enforcement purposes. “Real-time” remote biometric identification systems will be permissible in a restricted number of severe cases. Conversely, post remote biometric identification systems, where identification is conducted after a substantial delay, will only be authorized for the prosecution of serious crimes and solely following court approval.


   3. What about ChatGPT?


Generative AI models such as ChatGPT will not be categorized as high-risk but will be obligated to adhere to transparency requirements and EU copyright law. This entails:

  • Disclosing that the content was generated by AI.
  • Implementing measures to prevent the generation of illegal content.
  • Providing summaries of copyrighted data utilized for training purposes.


   4. Deep fakes


Deep fakes are now defined under the EU AI Act as “AI-generated or manipulated image, audio, or video content that resembles existing persons, objects, places, or other entities or events and would falsely appear to a person to be authentic or truthful”.


The finalized text of the EU AI Act outlines transparency requirements for providers and deployers of specific AI systems and general-purpose AI models (GPAI) that are more stringent than earlier drafts. These obligations include transparency mandates for deployers of deep fakes, with exceptions granted in cases where the use is authorized by law for detecting, preventing, investigating, and prosecuting criminal offenses.


In instances where the content constitutes an obviously artistic work, transparency obligations are limited to disclosing the presence of generated or manipulated content in a manner that does not impede the presentation or enjoyment of the artwork.


   5. What are the penalties for non-compliance?


Similar to the methodology employed under the European General Data Protection Regulation, fines for breaches of the Act will be calculated either as a percentage of the offending party’s global annual turnover in the preceding financial year or as a fixed sum, whichever is greater:

  • €35 million or 7% for infringements involving prohibited AI applications;
  • €15 million or 3% for violations of the Act's obligations; and
  • €7.5 million or 1.5% for the dissemination of inaccurate information.


Nevertheless, there will be proportional limits on administrative fines imposed on small and medium enterprises as well as start-ups.


   6. Next steps


The agreed-upon text is anticipated to be formally adopted in April 2024. It will become fully applicable 24 months following its entry into force, thus in 2026, but certain provisions will come into effect sooner:

  • The prohibition of AI systems presenting unacceptable risks will take effect six months after entry into force.
  • Codes of practice will be enforceable nine months after entry into force.
  • Regulations pertaining to general-purpose AI systems, which must adhere to transparency requirements, will be applicable 12 months after entry into force.
  • High-risk systems will be afforded additional time to comply with the requirements, as obligations concerning them will become applicable 36 months after entry into force.
by Aurilex 18 September 2026
A new EU-wide intellectual property right now protects the geographical names of craft and industrial products, strengthening the link between regional know-how, reputation and commercial value.
by Aurilex 27 August 2026
Evidence in French IP Infringement Proceedings: A Purchaser’s Lack of Independence No Longer Automatically Invalidates a Purchase Report
by Aurilex 12 August 2026
France’s Highest Court Draws a Clear Line Between Product Packaging, Commercial Names and Alcohol Advertising under the Law Evin.
by Aurilex 5 August 2026
The French Court of Cassation provides important guidance on the assessment of added subject-matter and intermediate generalisations in European divisional patents.
by Aurilex 10 July 2026
EU General Court reinforces the protection of trade marks with reputation The Obelix case
by Aurilex 2 June 2026
How the CJEU's landmark BSH v Electrolux ruling may allow the UPC to extend its reach to UK patents and reshape cross-border patent litigation in Europe.
by Aurilex 16 February 2026
On 3 December 2025, the French Court of Cassation issued an important ruling concerning the prohibition of added subject-matter in the context of a European divisional patent relating to a Wii video game console controller. The decision provides valuable guidance on: the assessment of inadmissible intermediate generalisations; the interpretation of patent documents by national courts; and the role of the “person skilled in the art” in the added subject-matter analysis under the EPC. Legal Framework Article 123(2) EPC prohibits amendments that extend the subject-matter of a European patent beyond the content of the application as originally filed. Under Article 138(1)(c) EPC, a European patent may be declared invalid if: its subject-matter extends beyond the content of the application as filed; or where granted on the basis of a divisional application, it extends beyond the content of the earlier (parent) application. The applicable test is whether the amended subject-matter derives directly and unambiguously , using common general knowledge, from the earlier application as filed. Background of the Case Nintendo Co. Ltd., proprietor of European divisional patent EP 518 relating to the Wii console controller, initiated enforcement actions against Bigben Interactive, a company marketing video game accessories. In response, Bigben brought revocation proceedings before the French courts against the French part of EP 518. During prosecution before the EPO, the patent claims had been amended to include the presence of an acceleration sensor located inside the controller housing. The Paris Court of Appeal (21 April 2023) invalidated the French part of the patent, holding that: the addition of the acceleration sensor resulted in an inadmissible intermediate generalisation, and the subject-matter extended beyond the content of the earlier application. Nintendo appealed to the Court of Cassation. The Court of Cassation’s Ruling The Court examined two central issues. 1. Distortion of the Patent Document The Court of Appeal had considered that Figure 8 and the description established a “necessary link” between: the acceleration sensor (68), and the processor (66). It concluded that isolating the acceleration sensor without expressly including a processor amounted to extracting a feature disclosed only in combination. However, the Court of Cassation held that this interpretation distorted the patent document. The parent application stated that: “ the acceleration sensor 68 can be used in combination with the processor 66 (or any other processor) to determine the inclination, altitude or position of the housing .” The wording did not establish an inextricable link between the acceleration sensor and a specific processor. By reading such a mandatory connection into the text, the Court of Appeal misrepresented the content of the earlier application and violated the principle prohibiting courts from distorting written evidence. This constitutes a significant reminder that national courts must apply a faithful and technically accurate reading of patent documents. 2. Failure to Define the “Person Skilled in the Art” The Court of Appeal further held that: there was no evidence that a video game controller necessarily included a processor; therefore, the addition of the acceleration sensor without explicitly including a processor constituted an inadmissible intermediate generalisation. The Court of Cassation rejected this reasoning. It recalled that the added subject-matter test requires assessing whether the amended feature derives directly and unambiguously from the earlier application from the perspective of the person skilled in the art, taking into account implicit technical knowledge. Crucially, the Court of Appeal failed to define the relevant skilled person and to examine whether, for such a person, an acceleration sensor would implicitly require data processing by a processor. If, from a technical standpoint, a processor is inherently necessary to process acceleration data, its presence may be implicit and need not be explicitly restated in the claim. By omitting this analysis, the Court of Appeal deprived its decision of a legal basis under Articles 123(2) and 138(1)(c) EPC. Practical Implications This decision is noteworthy for several reasons: 1. Reinforcement of the “Direct and Unambiguous” Test The ruling aligns French case law more closely with established EPO jurisprudence on added subject-matter. 2. Central Role of the Skilled Person Courts must explicitly define the relevant skilled person and assess implicit technical features through that lens. 3. Limits of the Intermediate Generalisation Doctrine Not every extraction of a feature from a disclosed embodiment amounts to an inadmissible intermediate generalisation. The decisive question is whether the feature is structurally and functionally linked to other elements in a manner that is technically inextricable. 4. Drafting and Prosecution Strategy for Divisional Applications For patent applicants, the decision highlights the importance of: drafting fallback positions with clear technical independence of features; explicitly addressing whether components are structurally mandatory or functionally optional; anticipating potential added subject-matter challenges in national revocation actions. Conclusion The Court of Cassation’s ruling provides welcome clarification on the added subject-matter analysis in the context of divisional patents. By emphasising both the prohibition of document distortion and the necessity of properly defining the person skilled in the art, the Court reinforces a technically grounded and EPC-consistent approach to invalidity assessments. The decision will likely influence future French revocation proceedings involving intermediate generalisations and divisional patents. Elsa Duboin, Trainee Lawyer Christine Chai, Managing Partner, Attorney-at-Law
by Aurilex 30 January 2026
Aurilex is pleased to announce that its Managing Partner, Christine Chai, has been recognised in the WTR 1000 – France 2026. The WTR 1000 is a leading international guide dedicated to identifying the world’s foremost trademark professionals, based on extensive research, peer reviews and client feedback. This recognition reflects our expertise in trademark law and broader intellectual property matters, as well as the trust placed in Aurilex by its clients. It also highlights the collective work and high standards of the Aurilex team in assisting brands with trademark protection, portfolio management and IP strategy in complex international and cross-border contexts, particularly across Europe. Aurilex regularly advises international clients on the development, protection and enforcement of their IP rights, combining legal precision with a strong understanding of multi-jurisdictional and strategic considerations.
by Aurilex 27 January 2026
When “neutral hosting” ends: Airbnb held outside the hosting safe harbour
by Aurilex 4 September 2025
France’s 2019 PACTE law introduced new administrative procedures for trademark invalidity and revocation , which came into force on 1 April 2020. This reform was a cornerstone of the modernization of industrial property enforcement: it aimed to make trademark law more accessible, faster, and less costly. The objectives were clear: to give businesses, especially SMEs, the tools to defend their rights more efficiently, to clear the register of unused or abusive marks, and to prevent trademarks from unfairly blocking market access or harming the public interest. Five years later, the French IP Office (INPI) has published its assessment in the report “ Cinq ans déjà ! Les procédures de nullité et de déchéance des marques devant l’INPI ”. The results show a system that has been widely adopted and is now a stable part of the French trademark landscape. A Procedure Embraced by Businesses Since 2020, nearly 2,200 applications for invalidity or revocation have been filed before the INPI, with more than 1,800 decisions delivered. The annual average—around 440 cases —is well above the level seen before judicial courts prior to the reform. SMEs are major users of the system, filing over 35% of applications, while foreign companies have also relied heavily on it. On the defence side, 38% of respondents are SMEs and 37% are individuals. Although professional representation is not mandatory, almost 90% of applicants are assisted by counsel. Lawyers are present in 51% of cases, industrial property attorneys in 41%. Stable and Efficient Litigation Framework Applications are split between invalidity (60%) and revocation (40%) . Within invalidity, 70% are based on relative grounds such as earlier trademarks, while absolute grounds like lack of distinctiveness or bad faith remain less common. Bad faith is a notable trend : more than 150 decisions have been issued on this basis since 2020, with a success rate of around 50%. Fraudulent filings are alleged in over 20% of invalidity actions. For revocation, non-use remains the overwhelming ground , though around 30 cases have also concerned degeneration or deceptive use. Duration and Outcomes The average duration of proceedings is 8.5 months . Uncontested cases can be resolved in 7 months, while complex disputes may last up to 17 months. The results are decisive: in 85% of cases, applications are deemed justified . More than 1,000 trademarks have been annulled or revoked since 2020. Around 30% of cases are closed following amicable settlements, and only 3% end in inadmissibility. Relative grounds: Over 85% of prior trademark claims succeed , including claims based on well-known marks. Company names: Only half of claims succeed, while other signs (domain names, trade names) often fail due to lack of proof or legal basis. Absolute grounds: Just 30% succeed (excluding bad faith), often because of evidentiary challenges . The INPI also rules on costs: in over half of the cases, the losing party is ordered to pay, with an average cost award of €680. Conclusion: A Maturing and Trusted System Five years after its introduction, the administrative procedure for invalidity and revocation has reached maturity. The high level of adoption, particularly by SMEs, and the strong success rate of actions confirm its effectiveness. The INPI’s system has become a trusted and efficient alternative to judicial litigation— faster, less costly, and widely accessible . For businesses navigating trademark strategy in France, it is now an indispensable tool. Clara Courret, Trainee Lawyer Christine Chai, Managing Partner, Attorney-at-Law