EU AI Act Passed, Key High‑Risk Requirements Delayed to 2027‑2028

The European Union’s Artificial Intelligence Act has completed the legislative process and entered into force on 2 August 2024, with most substantive obligations slated for 2 August 2026. The Act bans certain “prohibited” AI systems from 2 February 2025 and classifies AI into four risk categories, imposing the strictest requirements on high‑risk and general‑purpose systems.

In May 2026, the European Parliament and Council agreed to postpone several high‑risk obligations: biometric identification and recruitment‑screening compliance is now due by December 2027, safety‑critical component compliance by August 2028, and AI‑generated‑content watermarking by December 2026. Transparency duties, requiring clear user disclosures when interacting with AI, will start on 2 August 2026, supported by draft guidance from the European Commission.

The amendments give companies additional time to assess compliance, especially for sectors such as insurance, hiring, law enforcement, defence contracting, and education, while also introducing a ban on “nudifier” applications from December 2026. Businesses are advised to benchmark current transparency practices against the new guidance and to prepare governance frameworks for the delayed high‑risk obligations.