Ask the question. Get the evidence behind the answer.
Thembi Ask
Ask anything about EU policy.
Compare dossiers, detect recent signals and understand what matters for policy teams.
As of 23 September 2026, the AI Act (Regulation (EU) 2024/1689) is in its main implementation phase [1]. Its framework has been amended by Regulation (EU) 2026/1744, which revised the timetable and certain classification rules for high-risk AI systems [2].
1. State of play: implementation and revised timeline
The AI Act generally became applicable on 2 August 2026, although several categories of obligations apply on different dates [1]. The 2026 amendments revised important parts of the high-risk AI timetable [2].
| Requirement | Relevant application date |
|---|---|
| Most prohibited practices | 2 February 2025 [1] |
| Additional prohibited practices | 2 December 2026 [2] |
| High-risk AI — Annex III | 2 December 2027 [2] |
| High-risk AI — Annex I | 2 August 2028 [2] |
2. Key simplification measures
- Refined high-risk classification: clearer rules for AI integrated into regulated products [2].
- Non-safety uses: optimisation and quality-control functions are not automatically safety components [2].
- Health and safety safeguard: systems can still qualify as high-risk where malfunction could endanger people [2].
The Commission also launched the AI Act Service Desk and Single Information Platform to provide practical implementation support [3].
3. Guidance and enforcement
The European AI Office, the AI Board and national competent authorities coordinate implementation [4]. Policy teams should monitor final guidance on high-risk classification, Article 50 transparency duties, harmonised standards and common specifications [5][7].
4. What a policy team should monitor next
- 2 December 2026: additional prohibited practices and transitional requirements.
- 2 December 2027: high-risk obligations for Article 6(2) and Annex III systems.
- 2 August 2028: high-risk obligations for Article 6(1) and Annex I systems.
- Related files: the Cloud and AI Development Act and Cybersecurity Act 2 [8][9].
Overall, the direction is a more staged implementation with revised timelines, narrower classification rules in certain cases, and additional guidance and support [1][2].
Sources
- AI Act — European Commission
- Regulation (EU) 2026/1744 — EUR-Lex
- AI Act Service Desk and Single Information Platform
- European AI Office
- Draft guidelines on high-risk AI systems
- Consultation on high-risk classification
- Article 50 transparency guidance
- Cloud and AI Development Act — Legislative Observatory
- Cybersecurity Act 2 — Legislative Observatory
Thembi Ask helps your team explore complex policy, institutional and regulatory questions using our trusted knowledge base. Attach your own files, investigate a topic in depth and turn findings into structured PDF reports - without losing track of the sources behind every answer.
- 01Grounded answers
- 02Bring your own files
- 03Deep research
- 04PDF reports