What are the EU AI Act dates that matter for a deployer after the Digital Omnibus?
Five dates matter for a deployer: 2 February 2025 for Article 4 and the prohibited practices, 2 August 2026 for Article 50 transparency and for national supervision of Article 4, 2 December 2026 for the end of the marking grace period and two new prohibited practices, 2 December 2027 for most Annex III high-risk obligations, and 2 August 2028 for AI embedded in Annex I regulated products. Regulation (EU) 2026/1744, the Digital Omnibus on AI, set or moved the last three of those and left the first two alone.
The five dates
1. 2 February 2025: Article 4 and Article 5 applied. The AI literacy duty and the prohibited practices have applied since this date. The Omnibus rewrote the wording of Article 4 from 27 July 2026 but did not change when it applies. If your staff use an AI tool for work, you have been in scope for over a year and a half.
2. 2 August 2026: Article 50 transparency, and national supervision of Article 4. From this date, people interacting with a chatbot must be told they are dealing with AI unless it is obvious, and deepfakes and certain AI-generated content must be disclosed. The Omnibus did not defer these duties. The Commission's Q&A on AI literacy also gives this as the date from which national market surveillance authorities supervise and enforce Article 4. In Ireland, the Regulation of Artificial Intelligence Act 2026, signed on 21 July 2026, put that framework in place.
3. 2 December 2026: the marking grace period ends, and two new prohibitions apply. The one Article 50 concession in the Omnibus was for providers of generative AI systems already on the market before 2 August 2026. They have until 2 December 2026 to add machine-readable marking to synthetic output. Systems placed on the market on or after 2 August 2026 got no grace period. This is a provider duty, but if you deploy such a system, ask your vendor where they stand. The Omnibus also added two prohibited practices to Article 5 from the same date: AI that generates sexually explicit images, video or audio of an identifiable person without their consent, and AI that generates child sexual abuse material.
4. 2 December 2027: most Annex III high-risk obligations. The Omnibus moved the application date for most high-risk systems listed in Annex III, which includes AI used in employment, recruitment, education, credit and essential services, to 2 December 2027. For a deployer, that is when the Article 26 duties bite in full: using the system according to its instructions, assigning human oversight to people with the necessary competence, training and authority, and the rest. The Commission states that the training obligation for high-risk deployers remains in place. Under Article 111(2), a system already in service before that date is only brought in once its design changes significantly, and systems intended for use by public authorities must comply by 2 August 2030 in any case.
5. 2 August 2028: AI embedded in Annex I regulated products. The high-risk rules for AI that is a safety component of, or is itself, a product covered by Section A of Annex I, such as a medical device, now apply from 2 August 2028. The Omnibus moved machinery to Section B, so AI in machinery is regulated through the Machinery Regulation instead, which must carry matching AI requirements by the same date.
What did not move
Article 4, the existing Article 5 prohibitions and Article 50 did not move, apart from the marking grace period, and neither did the general-purpose AI model obligations on providers, which applied from 2 August 2025. If a summary of the Omnibus tells you the AI Act was delayed, the accurate version is that the high-risk chapters were delayed and the deployer duties that apply to almost every organisation were not.
What a deployer should do at each point
- Now: inventory the AI in use, run a literacy measure for the people who use it, and record both. That is Article 4, and it is what determines whether you are in scope for everything else.
- Before December 2026: ask any generative AI vendor you rely on about marking.
- During 2027: identify anything in your inventory that screens, scores or ranks people, and plan the Article 26 oversight roles and training before 2 December 2027.
Sources
- Regulation (EU) 2026/1744, the Digital Omnibus on AI, published 24 July 2026, in force 27 July 2026, for the revised application dates and the marking grace period.
- Regulation (EU) 2024/1689, the AI Act, Articles 4, 5, 26, 50 and 113.
- European Commission, AI literacy: questions and answers, for the supervision date and the high-risk training passage.
- Department of Enterprise, Tourism and Employment: AI Office of Ireland established, 30 July 2026.
This is general information about the EU AI Act, not legal advice. Enforcement is proportionate and case by case, and your sectoral regulator may issue guidance that applies to you.
