EU defers AI Act high-risk duties to December 2027

The Council of the EU gave its final green light to the AI Act simplification omnibus on 29 June, following the European Parliament's endorsement on 16 June. The package enters into force on the third day after publication in the Official Journal, expected in July — just ahead of the original 2 August 2026 compliance deadline it was racing to beat.

The headline change: obligations for standalone high-risk AI systems under Annex III — covering areas such as biometrics, critical infrastructure, education, employment and recruitment — move from 2 August 2026 to 2 December 2027, a sixteen-month deferral. Obligations for Annex I high-risk systems embedded in regulated products (medical devices, machinery, vehicles) shift from 2 August 2027 to 2 August 2028. The EU's stated reason is delayed designation of national competent authorities and unfinished harmonised standards.

This is relief, not repeal — and not everything moves. The Article 50 transparency obligations largely stay on the original schedule, meaning duties to disclose when users are interacting with an AI system still apply from 2 August 2026, and the grace period for watermarking obligations on existing systems runs only to 2 December 2026.

If you deploy AI in the EU market, act on two tracks. First, make sure any customer-facing chatbot or AI interaction discloses itself before August. Second, use the deferral productively: inventory your systems against Annex III (recruitment screening and employee-monitoring tools are common surprises), and build the risk-management and documentation processes now rather than in a rush during 2027.