EU AI Act deadlines: what really starts and what was postponed
The Digital Omnibus on AI, in force since July 2026, moved several of the most quoted dates. Here is the current calendar, separating what is already due from what lands later.
Updated on 25 July 2026
2 February 2025
Prohibited practices
The Article 5 bans apply: harmful manipulation, exploitation of vulnerabilities, social scoring and the other prohibited uses.
2 February 2025
AI literacy
Duty to support the AI competence of people operating these systems on your behalf. Softened by the Digital Omnibus, but in force from this date.
2 August 2025
General-purpose models, governance and penalties
Obligations for GPAI model providers kick in, along with the European governance structure and the penalty framework.
2 August 2026
Article 50 transparency obligations
Chatbots to disclose, synthetic content to mark, deepfakes and public-interest text to label. This is the deadline that touches most websites.
2 August 2026
Full application and powers of the authorities
The regulation becomes fully applicable and national authorities gain full powers. In Italy: AgID as notifying authority, ACN for market surveillance and penalties.
2 December 2026
Machine-readable marking for systems already on the market
Generative systems placed on the market before 2 August 2026 have until this date to meet Article 50(2). Deferral introduced by the Digital Omnibus.
2 December 2026
End of the transition on the new prohibition
The transitional period ends for the ban on systems generating non-consensual intimate material or child sexual abuse material.
2 August 2027
National regulatory sandboxes
Member States must have sandboxes up and running. Pushed back by one year by the Digital Omnibus.
2 December 2027
High-risk systems under Annex III
Obligations for stand-alone high-risk systems move to this date. It used to be 2 August 2026: anyone still quoting that date is working from outdated information.
2 August 2028
High-risk systems under Annex I
Obligations for AI embedded in products already covered by other EU harmonisation legislation.
Do you need to meet the transparency obligations?
The transparency kit lets you declare what you do and prepares the notices, the statement and the summary of the duties that follow. All in your browser, nothing sent anywhere.
Open the AI Act transparency kitSources
Frequently asked questions
›What exactly changes on 2 August 2026?
The Article 50 transparency obligations become applicable and the regulation enters full application, with national authorities gaining full powers. In Italy, AgID is the notifying authority and ACN handles market surveillance and penalties.
›Is it true that high-risk systems start on 2 August 2026?
Not any more. The Digital Omnibus on AI, adopted by Parliament on 16 June 2026 and by the Council on 29 June 2026, moved Annex III high-risk obligations to 2 December 2027 and Annex I ones to 2 August 2028. Plenty of articles online still quote the old dates.
›What are the penalties for non-compliance?
Up to 35 million euro or 7 per cent of worldwide turnover for prohibited practices, 15 million or 3 per cent for other obligations, and 7.5 million or 1 per cent for supplying false information to authorities. For SMEs and startups the lower of the two amounts applies.
›Do I need to do anything if I only use ChatGPT to write?
It depends on what you publish. The duty to disclose AI-generated text covers content published to inform the public on matters of public interest, and it does not apply where there is human editorial review with editorial responsibility. The AI literacy duty, on the other hand, applies to anyone using these tools at work.