Summary: The Act entered into force in 2024 but applies in four staggered waves from February 2025 to August 2027, with review obligations running to 2031.
Sources: raw/AI-regulations/Regulation - EU - 2024_1689.md (Articles 112, 113); raw/Laws/Enforcement of Chapter V under the EU AI Act...md; raw/Laws/The EU AI Act's Transparency Rules...md
Last updated: 2026-07-28
Art 113 — the staged application
The Regulation “shall enter into force on the twentieth day following that of its publication” (published 12 July 2024) and “shall apply from 2 August 2026.” However (source: Regulation - EU - 2024_1689.md, Art 113):
| Date | What applies |
|---|---|
| 2 Feb 2025 | Chapters I and II — scope, definitions, AI literacy, and Article 5 prohibitions |
| 2 Aug 2025 | Chapter III Section 4 (notifying authorities and notified bodies), Chapter V (GPAI), Chapter VII (governance), Chapter XII (penalties, except Art 101), and Article 78 (confidentiality) |
| 2 Aug 2026 | Everything else — including Chapter III high-risk requirements, Art 15, Art 50 transparency, and Art 73 incident reporting |
| 2 Aug 2027 | Article 6(1) high-risk systems — safety components of products under Annex I harmonisation legislation — and their corresponding obligations |
Two wrinkles worth holding on to
1. GPAI obligations bind a year before the Commission can enforce them. Chapter V applies from 2 August 2025, but Commission “supervision and enforcement powers against GPAI model providers will only come into force on 2 August 2026” (source: Enforcement of Chapter V…md). See enforcement-and-penalties.
2. An Art 50(2) extension exists but is provisional. “The AI Omnibus provisional agreement of May 2026 grants generative AI systems already on the market before that date until 2 December 2026 to meet the machine-readable marking requirement under Article 50(2)” (source: The EU AI Act’s Transparency Rules…md). Provisional agreement, marking requirement only, pre-existing systems only. Verify before relying on it.
Deadlines the Commission and AI Office were given
| Due | Obligation |
|---|---|
| 2 May 2025 | Codes of practice ready (Art 56(9)) |
| 2 Aug 2025 | Commission guidance on serious incident reporting (Art 73(7)); fallback for codes of practice via implementing act (Art 56(9)) |
| 2 Feb 2026 | Guidelines on high-risk classification with practical examples (Art 6(5)) |
Review and evaluation obligations
- Annually from entry into force: assess the need to amend Annex III and the Art 5 prohibited-practices list, reporting to Parliament and Council.
- By 2 Aug 2028, then every four years: evaluate the need for new Annex III headings, amendments to the Art 50 list, and improvements to supervision and governance.
- By 2 Aug 2028: evaluate whether the AI Office has sufficient powers, competences and resources.
- By 2 Aug 2028, then every four years: report on standardisation for energy-efficient GPAI development.
- By 2 Aug 2028, then every three years: evaluate voluntary codes of conduct for non-high-risk systems.
- By 2 Aug 2029, then every four years: full evaluation and review of the Regulation, including “an assessment with regard to the structure of enforcement and the possible need for a Union agency.”
- By 2 Aug 2031: assessment of enforcement over the first years of application, again addressing “the need for a Union agency.”
(source: Regulation - EU - 2024_1689.md)
The recurring “need for a Union agency” question in both the 2029 and 2031 reviews signals the Commission/AI Office arrangement is explicitly provisional in the legislature’s mind.
Related bodies appointed
The Scientific Panel and Advisory Forum were both appointed on 1 June 2026 — the Panel with up to 60 experts, the Forum with 174 members from over 700 applications (sources: raw/Laws/The Scientific Panel of Independent Experts...md, raw/Laws/The Advisory Forum...md). See governance bodies.