Deadline tracker
EU AI Act deadlines after the Digital Omnibus
The Digital Omnibus on AI moved the high-risk deadlines back and reinstated registration for systems that use the Art. 6(3) derogation. Prohibitions and GPAI duties are unchanged. Every countdown inside V77 Trust reads from this same table.
| Obligation | Articles | Original date | Applies from | Status |
|---|---|---|---|---|
Prohibited AI practices Social scoring, manipulative techniques, untargeted facial-image scraping, emotion recognition at work and in education, and the other Art. 5 practices are banned. The Omnibus adds two bans from 2 December 2026: AI systems that generate non-consensual intimate imagery, and AI systems that generate child sexual abuse material. Who: Everyone placing on the market, putting into service or using AI in the EU. | Art. 5 | 2 Feb 2025 | 2 Feb 2025 | In force |
General-purpose AI model obligations Technical documentation, copyright policy and training-content summary for GPAI providers; extra duties for systemic-risk models. Who: Providers of general-purpose AI models. | Arts. 51–55 | 2 Aug 2025 | 2 Aug 2025 | In force |
Transparency duties for deployers Deployers must disclose deep fakes, AI-generated public-interest text, and emotion-recognition or biometric-categorisation use. Who: Deployers of chatbots, generative AI and biometric systems. | Art. 50(3)–(4) | 2 Aug 2026 | 2 Aug 2026 | In force |
Machine-readable marking of AI-generated content Providers must mark synthetic audio, image, video and text outputs in a machine-readable, detectable way. This has applied since 2 August 2026; generative systems already on the market before then have until 2 December 2026. Who: Providers of AI systems that interact with people or generate content. | Art. 50(2) | 2 Dec 2026Omnibus | 23 days left | |
National AI regulatory sandboxes Each Member State must have at least one AI regulatory sandbox operational. Who: Member State authorities; innovators wishing to test under supervision. | Art. 57 | 2 Aug 2027Omnibus | 266 days left | |
High-risk AI — standalone systems (Annex III) Risk management, data governance, Annex IV documentation, logging, human oversight, accuracy and robustness, conformity assessment, registration, FRIA for certain deployers. Who: Providers and deployers of AI used in employment, credit scoring, insurance pricing, education, critical infrastructure, essential services, law enforcement and similar Annex III areas. | Arts. 6(2), 8–27, 43, 47–49, Annex III | 2 Dec 2027Omnibus | 388 days left | |
Registration of Annex III systems assessed as not high-risk The Commission proposed removing this duty; the final Omnibus text keeps it. A provider that relies on the Art. 6(3) derogation must still register the system in the EU database with its justification. Who: Providers relying on the Art. 6(3) derogation. | Arts. 6(3)–6(4), 49(2) | 2 Dec 2027Omnibus | 388 days left | |
High-risk AI — embedded in regulated products (Annex I) High-risk requirements for AI that is a safety component of products such as medical devices, machinery or toys. Who: Manufacturers and providers under EU product-safety legislation. | Art. 6(1), Annex I | 2 Aug 2028Omnibus | 632 days left |
Dates include the Digital Omnibus on AI, Regulation (EU) 2026/1744, in force since 27 July 2026. Not legal advice.
Free check · no login
Article 50 transparency check
Five yes/no questions. Your answers stay in your browser — nothing is sent to us.
1. Does your AI system interact directly with people (chatbot, voice agent, assistant)?
2. Does it generate synthetic audio, images, video or text?
3. Do you deploy emotion recognition or biometric categorisation on people?
4. Do you publish AI-generated or manipulated image, audio or video that resembles real people, places or events?
5. Do you publish AI-generated text to inform the public on matters of public interest, without human editorial review?