EU Agrees on AI Act Following Marathon Trilogue Negotiations
Summary
On December 8, 2023, EU officials reached an agreement on the AI Act, the world’s first comprehensive legislation regulating AI. The trilogue (discussions between the European Parliament, Commission and Council) resulted in a political agreement following marathon negotiations that began on December 6 and culminated in a deal late on December 8. Although lawmakers reached an agreement, the final draft is not yet available because the provisions and technical details still need to be drafted and finalized. The EU Parliament’s overview of the agreement indicates that several sticking points were resolved and the Act will include: • Bans on systems imposing unacceptable risk, such as social scoring, user manipulation, untargeted scraping of facial images from the internet or CCTV footage and certain emotion recognition systems. • Strict obligations for high-risk systems, classified as high-risk due to their significant potential harm to health, safety, fundamental rights, environment, democracy and the rule of law, including a right for citizens to receive explanations about decisions based on highrisk AI systems that impact their rights. • New obligations, including on transparency, on general purpose AI, in line with a two-tiered approach, with more stringent obligations on high-impact models with systemic risk. • Limits on law enforcement use of biometrics with narrow exceptions. • Support for innovation, such as regulatory “sandboxes” and “real-world-testing.” • Fines up to 7% of global turnover for non-compliance. We anticipate the final draft will be available in early 2024, with a vote on the Act likely held in the spring before EU Parliamentary elections in June 2024. Akin’s preliminary coverage is available here.
