infinitelaws · EU AI Act · applicability timeline

What applies when

The AI Act does not kick in all at once — obligations apply in stages between 2025 and 2030, and the Digital Omnibus amendment has shifted several dates. Here is the full schedule in one place, linked to the affected provisions.

1. 1 août 2024applicable

   Entry into force

   The AI Act entered into force on the twentieth day following its publication in the EU Official Journal. Obligations phase in gradually from this date (Article 113).

   Affected provisions

   • Art. 113
2. 2 février 2025applicable

   General provisions and prohibited practices

   Chapters I and II became applicable: subject matter and scope, definitions (Article 3), AI literacy (Article 4) and the bans on unacceptable AI practices (Article 5) — e.g. manipulative techniques, social scoring or untargeted scraping of facial images.

   Affected provisions

   • Art. 1–4
   • Art. 5
3. 2 août 2025applicable

   General-purpose AI, governance and penalties

   Rules for general-purpose AI models (Chapter V), notified bodies (Chapter III Section 4), EU and national governance (Chapter VII), confidentiality (Article 78) and penalties (Articles 99–100) became applicable.

   Affected provisions

   • Art. 28–39
   • Art. 51–56
   • Art. 64–70
   • Art. 78
   • Art. 99–100
4. Today13 juillet 2026
   2 août 2026upcoming

   General date of application

   The rest of the Act applies — including transparency towards people (Article 50), regulatory sandboxes (Articles 57–63), market surveillance and enforcement (Chapter IX) and transitional provisions (Article 111). Obligations for high-risk systems are deferred by the amendment (see below).

   Affected provisions

   • Art. 7
   • Art. 40–49
   • Art. 50
   • Art. 57–63
   • Art. 71–77
   • Art. 79–98
   • Art. 101
   • Art. 102–110
   • Art. 111
   • Art. 112
5. 2 décembre 2026Ω Omnibusupcoming

   New prohibitions and AI-content marking

   Under the amendment, AI systems generating non-consensual intimate material (so-called nudifier apps) and CSAM are prohibited (Article 5(1), points (ba) and (bb)). The transition for machine-readable marking of AI content also ends — systems placed on the market before 2 August 2026 must meet Article 50(2) by this date (Article 111(4)).

   Affected provisions

   • Art. 5
   • Art. 50
   • Art. 111
6. 2 août 20272 août 2026Ω Omnibusupcoming

   National sandboxes; legacy general-purpose AI

   Deadline for at least one national regulatory sandbox to be operational — postponed by a year by the amendment (Article 57). General-purpose AI models placed on the market before 2 August 2025 must also comply (Article 111(3)).

   Affected provisions

   • Art. 57
   • Art. 111
7. 2 décembre 20272 août 2026Ω Omnibusupcoming

   High-risk systems (Annex III)

   Obligations apply for stand-alone high-risk AI systems under Annex III — biometrics, critical infrastructure, education, employment, public services, law enforcement, migration and justice (Chapter III Sections 1–3, including classification under Article 6).

   Affected provisions

   • Art. 6–27
   • annexe III
8. 2 août 20282 août 2027Ω Omnibusupcoming

   High-risk systems (Annex I)

   Obligations apply for AI systems that are safety components of products covered by the sectoral legislation in Annex I — e.g. machinery, toys, medical devices or vehicles (Article 6(1)).

   Affected provisions

   • Art. 6
   • annexe I
9. 2 août 2030upcoming

   Legacy systems of public authorities

   High-risk AI systems already used by public authorities before the Act became applicable must be brought into compliance (Article 111(2)).

   Affected provisions

   • Art. 111
10. 31 décembre 2030upcoming

    Large-scale EU information systems

    AI systems that are components of the large-scale EU information systems listed in Annex X and were placed on the market before 2 August 2027 must comply by the end of 2030 (Article 111(1)).

    Affected provisions

    • Art. 111
    • annexe X