Another step forward for the AI Act; preliminary political agreement reached in the European Parliament
The AI Act is designed to create a single legal framework for the use of artificial intelligence (AI) in the European Union. While there is agreement with regard to the “big-picture” goals of the legislation, i.e. ensuring a high degree of security when interacting with AI systems, certain aspects of the Act have long been the subject of controversial discussion. Compromise solutions have now been identified in this regard.
High-risk AI system
The AI Act defines an AI system as “software that is developed with one or more of the techniques and approaches listed in Annex I and can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with.”
AI systems falling within the critical areas and cases listed in Annex III to the Act are automatically categorized as “high-risk” systems. But a concession has now been obtained on this point so that AI systems would only be classified as “high-risk” if they pose a “significant risk.”
A risk is considered to be “significant” if it has a significant impact based on its severity, intensity, probability of occurrence and duration, and if it has the potential to harm an individual, multiple persons or a specific group of people. AI systems are also considered to be high-risk if they pose significant health risks.
AI-based medical devices
Accordingly, medical devices which are based on AI components will continue to be classified as high-risk AI systems even after these changes to the definition of “high-risk,” so that they will have to satisfy the special regulatory requirements for high-risk AI systems in accordance with the AI Act in addition to the requirements of the MDR.
The questions relating to the treatment of self-learning AI and the classification of the resulting product changes as significant changes in terms of the MDR, requiring a new conformity assessment, remain unresolved. It remains to be seen whether this problem will be identified and resolved in a timely manner. Hopefully this will be the case, as it would create legal certainty for the affected economic operators.
The text of the AI Act is currently being finalized. The Act is expected to take effect before the end of this calendar year (becoming binding 24 months after its entry into effect).
Recommended actions
Medical device manufacturers whose products feature AI components will be subject to both the Medical Device Regulation
back