The AI Act doesn't land all at once — it phases in through 2028. Here's what already applies, what comes next, and which dates were just postponed, at a glance.
Kept current — with sources and the date it was last checked.
Affects: Everyone — the staggered deadlines begin
Regulation (EU) 2024/1689 is in force. The individual obligations apply in stages over the following years.
Affects: Anyone using or providing AI
Eight prohibited AI practices are banned (e.g. social scoring, manipulative systems, emotion recognition at work). Plus an obligation to ensure staff have adequate AI literacy.
Affects: Providers of general-purpose AI models (GPAI); the supervisory structure
Providers of general-purpose AI models must meet transparency, copyright and risk obligations. The EU AI Office starts operating and the penalty framework applies.
Affects: Anyone generating AI content or running chatbots
AI-generated content and deepfakes must be labelled; users must be able to tell when they are interacting with AI. This deadline was not postponed.
Affects: Providers of generative image/video/audio AI
New ban on AI-generated non-consensual intimate imagery and abuse material. The four-month transition for the watermarking obligation also ends.
Affects: Deployers & providers of high-risk AI: biometrics, critical infrastructure, education, employment/HR, migration & border control
Full high-risk obligations apply: risk management, data quality, technical documentation, human oversight and conformity assessment.
ChangedPostponed by the Digital Omnibus (political agreement 7 May 2026): originally 2 August 2026 → now 2 December 2027.
Affects: Makers of products with embedded AI: machinery, toys, lifts, medical devices, etc.
High-risk obligations for AI embedded as a safety component in already-regulated products.
ChangedPostponed by the Digital Omnibus: originally 2 August 2027 → now 2 August 2028.
You know the deadlines — but not which risk class your system falls into? Find out in a few minutes.
Go to the risk checkLast checked: 21 June 2026
This timeline is a plain-language overview, not legal advice. The official text of the regulation prevails.
Partly. Since 2 February 2025 the prohibited practices and the AI-literacy obligation apply; since 2 August 2025 the obligations for general-purpose AI models (GPAI) plus governance and the penalty framework. The big high-risk obligations come later — and were recently postponed.
Yes. Under the “Digital Omnibus”, Council and Parliament agreed on 7 May 2026 to postpone the high-risk obligations — Annex III to 2 December 2027, Annex I to 2 August 2028. Formally the change only takes effect once published in the EU Official Journal, which is expected before 2 August 2026.
Two things matter regardless of the postponement: staff AI literacy (in force since February 2025) and the transparency obligation for AI-generated content (from 2 August 2026). If you're planning high-risk AI, the delay buys time — but it's not the all-clear.
We update it after every relevant change and show the date it was last checked, with sources, at the top. Precisely because a lot is moving right now, you'll find the latest state here.
We classify your AI systems, prioritise the deadlines that actually matter for you, and implement with you — pragmatically, without panic.