EU AI Act amendment delays high-risk deadlines

The EU’s Digital Omnibus on AI postponed the main high-risk requirements under the AI Act. Rules for uses listed in Annex III, including recruitment software, credit scoring and exam scoring, now start on 2 December 2027. Requirements for AI embedded in products covered by Annex I begin on 2 August 2028. Businesses should use the additional time to classify systems and prepare risk management, documentation, human oversight and conformity assessments.

Existing prohibitions and obligations for general-purpose AI models remain in force. The transparency duties of Article 50 apply from 2 August 2026, including informing people when they interact directly with AI. Generative systems placed on the market before 2 August 2026 have until 2 December 2026 to mark their output. A new prohibition covering systems that generate non-consensual intimate material or child sexual abuse material also applies from 2 December 2026.

AI literacy remains mandatory, but providers and deployers must now take measures to support staff rather than guarantee a particular proficiency level for every individual. The amendment also extends several simplified compliance measures to small mid-cap enterprises. US companies can fall within scope when supplying AI systems or general-purpose models to the EU market, or when their systems’ outputs are used in the EU. Compliance planning should include an inventory of systems, assigned business roles and records of literacy measures.