In force since 2 August 2026 · Last updated 29 July 2026
What has applied since 2 August 2026.
The Digital Omnibus on AI moved most high-risk obligations to 2 December 2027 (Annex III) and 2 August 2028 (Annex I). Article 50 transparency obligations, the existing prohibitions and GPAI duties were not deferred and are in effect today. The next milestone is 2 December 2026. Regulators expect real evidence - not intent.
Start free readiness assessmentWhat is in effect today
- Chatbot disclosure - users told when they're interacting with AI (Art. 50(1))
- AI-generated content labelling - deepfakes and synthetic media marked as such (Art. 50(4))
- Emotion recognition & biometric categorisation notices to affected persons (Art. 50(3))
- Machine-readable marking of AI-generated output for new systems (Art. 50(2)) - grace period for pre-existing systems runs until 2 Dec 2026
- Existing prohibitions, Article 4 competency support and GPAI obligations remain in force
What comes next
The bulk of high-risk obligations - the heaviest part of the AI Act - has moved. Use the extra runway; don't waste it.
2 Dec 2026
Next milestone: marking grace period ends for pre-existing generative systems, and the new Article 5 ban on AI nudifiers and AI-generated CSAM starts
2 Aug 2027
Member states must have AI regulatory sandboxes established
2 Dec 2027
Annex III high-risk systems (HR, credit, education, law enforcement, biometrics, essential services)
2 Aug 2028
Annex I high-risk AI embedded in regulated products (medical devices, machinery, vehicles, toys)
Frequently asked
Wasn't 2 August 2026 supposed to be THE deadline?
It was - until the Digital Omnibus on AI (Regulation (EU) 2026/1744, in force 27 July 2026) deferred the high-risk obligations to 2 Dec 2027 (Annex III) and 2 Aug 2028 (Annex I). The Article 50 transparency obligations were not deferred and have applied since 2 August 2026.
What has applied since 2 August 2026?
The transparency layer of the AI Act: chatbot disclosure, labelling of AI-generated content and deepfakes, notices around emotion recognition and biometric categorisation, and - for new systems - machine-readable marking of AI-generated output.
Does the deferral apply to us if we only use AI from external vendors?
The deferral applies equally to deployers of high-risk systems. But transparency obligations, existing prohibitions, Article 4 competency support and GPAI-related duties are unaffected - and vendor evidence still takes months to gather.
What are the fines for non-compliance?
Article 99 sets three tiers, capped at the higher of a fixed amount or a percentage of worldwide annual turnover: up to €35M or 7% for prohibited practices (Art. 99(3)), €15M or 3% for most other breaches - including high-risk obligations and Article 50 transparency duties (Art. 99(4)) - and €7.5M or 1% for supplying incorrect, incomplete or misleading information to authorities (Art. 99(5)). The SME rule and a case-specific assessment apply.
Is the Omnibus legally binding?
Yes. The Digital Omnibus on AI (Regulation (EU) 2026/1744) was signed on 8 July 2026, published in the Official Journal of the EU on 24 July 2026 and entered into force on 27 July 2026. The amended deadlines now apply across the EU.