EU AI Act
The EU AI Act applies in phases, not on one date.
Different obligations switch on at different times (Article 113). Here is the full schedule, with the deadlines the Digital Omnibus changed clearly marked. As of September 2026, everything up to August 2026 is already in effect; the next milestones, new prohibitions in December 2026 and the deferred high-risk deadlines, are still ahead.
Entry into force
- The Act formally enters into force and starts the clock for every deadline below.
- No operative obligations are triggered on this date itself.
Art. 113
Prohibited practices + AI literacy
- The Article 5 bans on unacceptable-risk AI take effect.
- The Article 4 AI-literacy duty applies to everyone who provides or uses AI.
- The first substantive duties, live now, whatever tier your systems fall into.
Art. 5 · Art. 4
GPAI models, governance, penalties
- General-purpose AI model obligations apply (Chapter V).
- The governance framework and national competent authorities are in place.
- The penalty regime becomes applicable.
Art. 51–56 · Art. 99
Transparency + most of the Act
- The Article 50 transparency duties apply (disclose AI interaction, label synthetic content, flag deepfakes).
- Most remaining provisions become applicable and enforcement begins.
- Annex III high-risk duties were originally due here, then deferred (see later).
Art. 50
New prohibitions + synthetic-content deadline
- New Article 5 prohibitions apply (added by the Digital Omnibus): non-consensual intimate deepfakes and child-sexual-abuse-material AI systems.
- Transition deadline for providers of synthetic-content generators already on the market to meet the Article 50(2) marking duty.
Art. 5 · Art. 50(2)
AI regulatory sandboxes
- Each Member State must have at least one AI regulatory sandbox operational at national level.
- A supervised space to develop, test and validate an AI system before it goes to market.
Art. 57
was 2 Aug 2026, deferred by the Digital Omnibus
Annex III high-risk obligations
- The full high-risk set for Annex III systems (hiring, credit, education, essential services, biometrics, and more).
- Provider requirements (Art. 9–15) and the Art. 16 provider duties; deployer duties (Art. 26); the fundamental-rights impact assessment (Art. 27).
- Deferred by the Digital Omnibus on AI (Reg (EU) 2026/1744).
Art. 9–15 · Art. 16 · Art. 26 · Art. 27
was 2 Aug 2027, deferred by the Digital Omnibus
Annex I product-safety high-risk
- High-risk obligations for AI that are safety components of products already regulated under EU product-safety law (medical devices, machinery, and similar).
- Gives those sectors time to align AI Act conformity with their existing sectoral regimes.
- Also deferred by the Digital Omnibus.
Art. 6(1) · Annex I
Scroll sideways to see the full timeline →
Transitional rules (easy to miss)
GPAI models already on the market
General-purpose AI models placed on the market before 2 August 2025 have until 2 August 2027 to come into compliance.
Art. 111(3)
High-risk systems already in use
High-risk systems placed on the market before their applicability date are generally caught only if they undergo a significant change in design after it.
Art. 111(2)
Dates rest on the amended Act
The high-risk deadlines above reflect the Digital Omnibus deferral. Confirm the current position with counsel before acting on a specific date.
Reg (EU) 2026/1744
What this means for you
Do not treat the AI Act as a single deadline. A company using an AI hiring tool already has AI-literacy and prohibited-practice duties live today, faces the full Annex III high-risk set from December 2027, and, if that tool is also a safety component of a regulated product, a further Annex I deadline in August 2028. The practical question is not “when does the Act apply” in the abstract, but which of these dates applies to your specific systems and role.
Find your dates, freeBased on the European Commission’s official implementation timeline for the EU AI Act (Regulation (EU) 2024/1689, Article 113), as amended by the Digital Omnibus on AI (Reg (EU) 2026/1744). Compliance support, not legal advice.