Hael
Sign inRequest a demo
EU AI Act · Requirements

EU AI Act timeline and key deadlines

Updated 30 June 2026 · 6 min read
Key takeaway
The EU AI Act does not switch on all at once. Its obligations phase in over several years, from the first prohibitions in February 2025 through to product-embedded high-risk rules in 2028. Some duties are already enforceable today, while the most consequential high-risk obligations were rescheduled in 2026. This guide sets out the key dates and flags what is provisional.
  • The Act phases in from February 2025 to 2028; it does not all apply at once.
  • Prohibited practices (Feb 2025) and GPAI rules (Aug 2025) are already in force.
  • The Digital Omnibus agreement of May 2026 deferred high-risk Annex III obligations to 2 December 2027 (provisional, pending formal adoption).
  • Early preparation is the sensible path, since buyers ask about readiness regardless of the legal date.
  • Current as of June 2026. This is general information, not legal advice.

The key dates

  • 1 August 2024: The Act entered into force, starting the phased clock.
  • 2 February 2025: Prohibited AI practices and AI literacy obligations became enforceable. Banned uses such as social scoring had to stop.
  • 2 August 2025: Obligations for general-purpose AI (GPAI) models took effect, including technical documentation, a training-content summary, and copyright measures. National authority designation and the penalties framework also applied from this point.
  • 2 December 2026 (provisional): Under the Digital Omnibus agreement, certain Article 50 transparency duties and two further Article 5 prohibitions are set to apply.
  • 2 December 2027 (provisional): The obligations for stand-alone high-risk AI systems under Annex III (such as hiring, credit, and biometrics) are scheduled to apply. This was deferred from the original date of 2 August 2026.
  • 2 August 2027: Providers of GPAI models that were already on the market before 2 August 2025 must be in full compliance by this date.
  • 2 August 2028 (provisional): High-risk AI embedded in regulated products under Annex I is scheduled to apply.

The important 2026 change

In May 2026, the EU reached a provisional agreement known as the Digital Omnibus, which deferred the high-risk Annex III deadline from 2 August 2026 to 2 December 2027, and Annex I product-embedded systems to 2 August 2028. This gives most organisations significant extra time on the heaviest obligations. However, this agreement was provisional and pending formal adoption at the time of writing, so the dates should be treated as the planning anchor while confirming final adoption later in 2026. The earlier-phase obligations (prohibited practices since February 2025 and GPAI since August 2025) are unchanged and remain in force.

What this means in practice

The phased timeline rewards organisations that prepare early rather than waiting for each deadline. Even with the high-risk date deferred to late 2027, building an AI inventory, classifying systems, and standing up governance now is the sensible path, because the work takes time and because enterprise buyers are already asking about readiness regardless of the legal deadline. The prohibitions and GPAI rules are live today, so those cannot wait.

A note on staying current

Because the timeline has already shifted once and parts remain provisional, treat any single date as a planning anchor and confirm against the official EU sources before relying on it. The direction of travel is settled even where specific dates move.

Key terms

Digital Omnibus
The May 2026 provisional EU agreement that deferred several EU AI Act deadlines.
Phased application
The Act's design of switching on different obligations at different dates.
High-risk deadline
The date by which stand-alone Annex III high-risk obligations apply (2 December 2027, provisional).
GPAI deadline
2 August 2025 for new models; 2 August 2027 for models already on the market before that date.
Prohibited practices
Banned AI uses under Article 5, enforceable since 2 February 2025.

References

Free check

See where you stand on EU AI Act, free.

Answer a few questions and get an indicative view of what EU AI Act expects of your AI systems and where you stand today — no sign-up to see your result.

Indicative, not legal advice.
EU AI Act · indicative readiness
HAEL FREE TOOL
Applicability
Applies to your AI use
What's expected
Risk classification · governance · documentation · oversight
Where you stand
Banded result · pointed to the gaps that matter most
Result
On-screen, free · optional PDF
Pre-scoped to EU AI Act~ 5 MIN