In the EU, the first steps of the legislative process on regulating AI process were completed on 13 March 2024, when the European Parliament approved the EU AI Act, and on May 22, when the EU Council approved it.

The AI Act seeks to enshrine safety and compliance with fundamental rights, whilst promoting innovation. It has extraterritorial reach and applies outside the EU to commercial providers placing on the market or putting into service AI systems in the EU, regardless of whether the providers are based within the EU.

It applies a risk-based approach to AI, introducing EU-wide requirements for AI systems, and uses a sliding scale of rules based on the risk of the AI system (unacceptable risk; high risk; and limited risk). It also introduces specific powers to impose high penalties for non-compliance of data requirements, breaching rules on general-purpose AI models and providing incorrect information, incomplete or misleading information.

Once the act becomes applicable, it will intersect with the EU medical devices regulatory regime (the MDR and IVDR). Software intended by a manufacturer to be used, alone or in combination, for a medical purpose such as diagnosis, prevention, monitoring, prediction, prognosis, treatment or alleviation of a disease is classified as a software medical device under the EU (and similarly under the the UK) regulatory system.

Underpinning the EU AI Act will be changes to product liability legislation to expand it to embrace AI and ML. “This is the first time we will be applying strict product liability law to services based on technology platforms, not products per se,” Ropes & Gray's Lincoln Tsang observed.

Moving Towards A Total Lifecycle Regulatory Approach For AI Devices

PCCPs On FDA And Global Agenda For 2024

More from Market Access

More from In Vivo