European Union AI Regulation

What the agreement does

The agreement defers the bulk of high-risk obligations to fixed dates: 2 December 2027 for standalone Annex III high-risk systems and 2 August 2028 for high-risk AI embedded in regulated products (Annex I). The Article 57 regulatory sandbox deadline is moved from 2 August 2026 to 2 August 2027. The Commission's originally proposed conditional trigger mechanism (linked to harmonised standards and supporting acts) has been dropped in favour of these fixed dates.

What still activates on 2 August 2026

Article 50 transparency obligations for new AI systems take effect on 2 August 2026 regardless of the Omnibus: chatbot disclosure, synthetic-content watermarking by providers, emotion-recognition and biometric-categorisation notices by deployers, and deepfake / AI-generated public-interest text disclosure. The Commission published a final, voluntary Code of Practice on the marking of AI-generated content on 10 June 2026 to support compliance with the watermarking obligation.

New December 2026 obligations from the Omnibus

The agreement adds two further dated obligations applying from 2 December 2026:

• Synthetic-content watermarking for providers of legacy systems placed on the market before 2 August 2026 (new systems placed on the market from 2 August 2026 are already in scope via Article 50).
• A new Article 5 prohibition on AI systems generating non-consensual intimate imagery and child sexual abuse material ("nudifier" systems), with a safe harbour for providers that deploy effective preventive safeguards.

Implementation timeline

1 August 2024: Entry into force. AI Act published in the Official Journal on 12 July 2024 and entered into force on 1 August 2024. AI literacy obligations under Article 4 began immediately.

2 February 2025: Prohibited practices apply. Ban on unacceptable-risk AI systems took effect: social scoring, subliminal manipulation, and most real-time remote biometric identification in public spaces. Commission guidelines on prohibited practices and the definition of an AI system published 6 February 2025.

2 August 2025: GPAI rules and governance. Obligations for general-purpose AI model providers became applicable. AI Office, European AI Board, Scientific Panel and Advisory Forum operational. Member states required to designate national competent authorities and set penalty regimes. GPAI Code of Practice published.

7 May to 16 June 2026 (current period): The Digital Omnibus on AI cleared its parliamentary stage. Provisional political agreement reached 7 May 2026, confirmed by COREPER 13 May 2026, and approved by the European Parliament in plenary 16 June 2026 (423 to 57, with 174 abstentions). Formal Council adoption is the only remaining step, expected before 2 August 2026, after which the amending regulation is published in the Official Journal and enters into force three days later.

2 August 2026: Article 50 transparency obligations apply to new AI systems regardless of the Omnibus: chatbot disclosure, synthetic-content watermarking, emotion-recognition and biometric-categorisation notices, and deepfake and AI-generated public-interest text disclosure. The Commission published a final, voluntary Code of Practice on the marking of AI-generated content on 10 June 2026. High-risk AI system deadlines, originally 2 August 2026 under the AI Act as adopted, are deferred by the Digital Omnibus to 2 December 2027 (Annex III) and 2 August 2028 (Annex I).

2 December 2026: Synthetic-content watermarking compliance date for providers of legacy systems placed on the market before 2 August 2026. A new Article 5 prohibition on AI systems generating non-consensual intimate imagery and child sexual abuse material applies from this date, with a safe harbour for systems that deploy effective preventive safeguards.

2 August 2027: Regulatory sandbox deadline. Under Article 57, each member state must have at least one AI regulatory sandbox operational. The Digital Omnibus defers this from 2 August 2026 to 2 August 2027.

2 December 2027: Fixed enforcement date for standalone high-risk AI systems (Annex III) under the Digital Omnibus.

2 August 2028: Fixed enforcement date for high-risk AI embedded in regulated products (Annex I) under the Digital Omnibus. Delegated acts under the Machinery Regulation incorporating AI Act high-risk requirements to apply by this date.

2 August 2029: Commission evaluation and report on AI Act functioning.

2030: High-risk AI in large-scale EU IT systems (SIS, VIS, Eurodac) must comply.

Article 6 guidelines

The Commission missed its 2 February 2026 deadline to publish guidelines on the practical application of Article 6 (high-risk classification), citing the need to integrate substantial stakeholder feedback. A revised set of guidelines is being prepared across 2026, covering high-risk classification, transparency, incident reporting, fundamental rights impact assessments, provider and deployer obligations, and the interplay of the AI Act with other EU legislation. The Commission ran a public consultation on the draft classification guidelines, which clarify how Articles 6(1) and 6(2) apply, and extended the consultation deadline from 23 June to 23 July 2026. Final guidelines remain expected by end of 2026.

Harmonised standards

The European standardisation bodies CEN and CENELEC failed to deliver harmonised technical standards for high-risk AI requirements by their August 2025 target. Standardisation work remains ongoing, with delivery now expected by end of 2026. The absence of harmonised standards is the principal technical rationale for the Omnibus delay. Under Article 40 AI Act, compliance with harmonised standards gives high-risk AI systems a presumption of conformity with the Chapter III Section 2 requirements; without those standards, providers lack the primary compliance route the AI Act was designed around.

National authority designations

Many member states have not yet formally designated their national competent authorities (market surveillance and notifying authorities) despite the 2 August 2025 deadline. As of early 2026, only three member states had completed full designation; approximately ten had pending legislative proposals; and fourteen had yet to act.