EU AI Act formally adopted by the Council

The Council of the European Union gave final approval to the AI Act, the last step in a legislative process that had run since the European Commission’s original 2021 proposal. The European Parliament had already adopted the text by 523 votes to 46 in March 2024; the Council’s sign-off on 21 May did not change the substance but converted political agreement into binding law, since under EU procedure a regulation needs approval from both bodies before it can be published.

The Act keeps the risk-tiered structure negotiators had settled on in December 2023: practices deemed an unacceptable risk — social scoring, most real-time biometric identification in public, emotion inference in workplaces and schools — are banned outright; “high-risk” systems in areas such as employment, credit and law enforcement face conformity assessment, documentation and human-oversight requirements; and general-purpose models face transparency obligations that scale with a compute threshold, with additional obligations for models deemed to carry “systemic risk.”

Adoption by the Council was not the same as the law taking effect. The regulation still required formal signature, translation into all EU official languages, and publication in the Official Journal — which followed on 12 July 2024 — before entering into force on 1 August 2024. Even then, most obligations phased in over the following two years: bans on prohibited practices applied from February 2025, obligations on general-purpose AI models from August 2025, and the bulk of high-risk system requirements not until August 2026. The Act became, on paper, the first comprehensive AI-specific statute passed by a major regulatory bloc, with the practical test of its enforcement still years off at the point of adoption.