The AI Act coming into force on 1 August 2024 marks a turning point in the European regulation of artificial intelligence. For innovative companies and intellectual property professionals, a crucial question arises: what are the real impacts of this new regulation on patent rights? Contrary to popular belief, the answer is more nuanced than it appears.
The AI Act: complementary, not substitutive, regulatory framework
The AI Act, officially called the« European Union Artificial Intelligence Act«constitutes the world's first comprehensive regulation on AI. However, it does not directly amend existing patent law. As the CNIL clarified in its initial Q&A, the AI Act «complements the GDPR without replacing it» – the same principle applies to intellectual property law.
The regulation primarily focuses on classifying AI systems according to their risk level, transparency and conformity obligations, as well as the governance and oversight of AI systems. It does not affect the fundamentals of patent law.
The conditions of patentability established by the’European Patent Office (OEB) remain identical. An invention must always present absolute novelty compared to the state of the art, an inventive step characterised by its non-obviousness, industrial applicability and technical character. These criteria, proven for decades, remain the cornerstone of patent protection.
The persistent challenges of AI patentability
The question of the artificial inventor: a debate definitively settled
The DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) case has decisively marked international jurisprudence. Dr Stephen Thaler had attempted to have his AI system recognised as the sole inventor of two innovations: a food container with a fractal structure and a light signalling device.
The decisions made by the principal courts are unanimous and final. The European Patent Office, in its decisions J 0008/20 and J 0009/20 of 21 December 2021, clearly established that only a natural person can be named as inventor. The British Supreme Court relied on the Patents Act 1977, which expressly requires the inventor to be a «natural person». For its part, the US Court of Appeals for the Federal Circuit, in the 2022 ruling of Thaler v. Vidal, held that the term «individual» can only refer to a natural person.
This consistent case law reflects a fundamental legal reality: AI systems, lacking legal personality, can neither hold rights nor transfer them.. This legal incapacity constitutes an insurmountable obstacle to their recognition as inventors within the meaning of the European Patent Convention.
Patentability criteria for AI-related inventions
According to the EPO Guidelines, an AI-related invention crosses the patentability threshold when it directly produces a technical effect such as ABS control or video compression, manages computing resources like memory or processor load, or is designed based on specific technical considerations.
The patent EP3989126 B1 granted by the EPO perfectly illustrates these criteria. This invention concerns a convolutional neural network for the automatic identification of radio-isotopes in gamma spectra. It was patented because it resolves a real technical problem in the medical field, applies AI in a concrete context, and precisely defines the architecture of the neural network used.
Indirect impacts of the AI Act on patent practice
Strengthening transparency requirements
The AI Act imposes substantial transparency obligations for high-risk AI systems, in accordance with its Article 13. This requirement may paradoxically facilitate the assessment of patentability by making innovations more understandable to patent examiners. Technical descriptions in patent applications are thereby improved, and establishing human authorship of inventions becomes easier.
This increased transparency is a double-edged sword for companies. On the one hand, it can strengthen the robustness of patent portfolios by providing richer and more precise documentation. On the other hand, it can reveal information that the company would have preferred to keep confidential as part of its competitive strategy.
Enhanced documentation and traceability
The documentation requirements imposed by the AI Act, detailed in its Article 11, are transforming how companies document their innovation processes. This enhanced documentation facilitates proof of inventive activity by tracking the creative process, establishes a precise timeline of technical developments, and irrefutably demonstrates human intervention in the invention process.
This development addresses growing concern among patent offices regarding the demonstration of human inventorship in AI-assisted inventions. The detailed documentation required by the AI Act can thus serve as tangible proof of this essential human contribution.
The disclosure dilemma
The transparency obligations imposed by the AI Act create a real strategic dilemma for innovative companies. On one hand, the regulation requires the disclosure of certain information about how AI systems work. On the other hand, patent protection sometimes necessitates maintaining secrecy before filing an application.
This tension forces businesses to rethink their intellectual property timelines. They must now nimbly balance regulatory compliance demands with their intellectual property strategy, anticipating opportune moments for patent filings and assessing the risks of premature disclosure.
Practical challenges for innovative companies
Cross-compliance: a major organisational challenge
Companies now operate in a complex legal environment where three regimes overlap. The AI Act imposes its classification of systems and risk assessments. Patent law maintains its patentability conditions and traditional protection strategies. The GDPR continues to govern the protection of personal data used by AI.
This regulatory overlap requires cross-functional governance involving legal, R&D, and compliance teams closely. Traditional organisational silos become counterproductive in this new context, where every technical decision can have multiple regulatory implications.
Protection strategies adapted according to the risk level
For high-risk AI systems, companies must anticipate transparency obligations when drafting their patents. This anticipation involves careful documentation of human intervention in innovation and a prior assessment of the impact of disclosure requirements on trade secrets.
Limited-risk AI systems offer more regulatory flexibility, allowing intellectual property protection to be optimised according to traditional strategies. Nevertheless, even in these cases, constant monitoring of evolving regulatory interpretations remains essential.
Case study: the difficulties of the Inari demand
The patent application WO2023250505 filed by the company Inari perfectly illustrates the current pitfalls of patentability in AI. This application concerns the use of «machine learning models» to predict the effects of genetic modifications on agronomic traits.
The analysis of this application reveals several recurring problems in the field. The claims use functional and general terms, making the scope of protection uncertain and difficult to assess. The procedural status of the application, «deemed withdrawn» by the EPO, reflects the difficulties the company has encountered in meeting patentability requirements.
This case demonstrates the crucial importance of drafting precise claims and clearly defining the technical character of AI-related inventions. It also highlights the need for specialised legal expertise to navigate the intricacies of AI patentability.
Future outlook and ongoing reforms
The product World Intellectual Property Organization (WIPO) is currently conducting in-depth consultations on the impact of AI on patent law. These efforts are exploring several avenues for reform, including adapting patentability criteria to AI-generated inventions, potentially creating new specific rights for AI creations, and harmonising international practices.
The European approach to regulating AI, embodied by the AI Act, could serve as a model for future harmonisation of intellectual property rules globally. The emphasis on transparency and human accountability in the AI Act is already influencing thinking on the evolution of patent law in other jurisdictions.
Strategic recommendations for the future
Companies must conduct a compliance audit to assess the impact of the AI Act on their existing patent portfolio. This assessment should be accompanied by training for innovation teams on the new regulatory requirements and adaptation of internal processes to integrate the AI Act's constraints into filing procedures.
Active legal monitoring is becoming indispensable for keeping up with the evolving interpretations and case law in this constantly changing field. Industrial property advisors must also adapt their practices by updating their drafting strategies to meet transparency requirements, assisting their clients in managing cross-compliance, and maintaining continuous training on regulatory and jurisprudential developments.
Towards a new balance between innovation and regulation
The AI Act does not revolutionise patent law, but it creates a complex regulatory environment that requires a completely rethought strategic approach. Innovative companies must now integrate the AI Act's compliance requirements into their intellectual property strategy, while continuing to adhere to traditional patentability criteria.
The main challenge lies in the intelligent management of transparency: how to meet the obligations of the AI Act without compromising the protection of your innovations? The answer lies in a coordinated approach between legal, technical, and compliance teams, supported by specialised expertise in this rapidly evolving field.
In this new regulatory landscape, the quality of documentation and the demonstration of human intervention are becoming key elements for securing your intellectual property rights. The AI Act, far from being an insurmountable obstacle, can therefore become a lever for strengthening your protection strategies, provided you approach it with the right tools and the right legal expertise.