EU AI Act / Key dates
What applies now, and what is still being finalised
The AI Act phases in over several years, and one major deadline is currently moving. Here is the honest picture: what is settled, what is confirmed, and what is still pending.
Current as of July 2026.
Already in force
Settled, and not moving
- 1 August 2024in force
The law entered into force
The EU AI Act became law across all 27 member states. As a regulation, it applies directly, with no national laws needed to bring it in.
- 2 February 2025in force
Banned practices and AI literacy went live
The short list of outright-banned AI uses took effect, and so did the duty to keep any staff who use AI at work basically literate in how it works and where it fails.
- 2 August 2025in force
General-purpose AI rules and governance began
Duties for the general-purpose models behind tools like ChatGPT started, together with the EU bodies that oversee the Act and the framework for penalties.
The full timeline
2024 to 2028, milestone by milestone
Every milestone below carries a status: in force today, a confirmed future date, or agreed but pending official publication. Note that stand-alone high-risk systems move to December 2027, not 2026. That distinction is often confused, and it matters.
- 1 August 2024in force
The AI Act enters into force
The regulation becomes EU law. The countdown to each set of duties starts here.
- 2 February 2025in force
Banned AI practices and the AI literacy duty apply
A short list of AI uses is prohibited outright, and any organisation whose staff use AI at work must keep them basically literate in how it works and where it fails.
- 2 August 2025in force
Rules for general-purpose AI models, plus governance and penalties, apply
Duties begin for the general-purpose models behind tools like ChatGPT, along with the EU bodies that oversee the Act and the framework for penalties.
- 2 August 2026confirmed
Transparency duties apply (chatbots and AI-generated content)
People must be told when they are dealing with AI or looking at AI-generated content. Confirmed, and not delayed by the Digital Omnibus reform.
- 2 August 2026confirmed
Enforcement powers for the AI literacy duty and general-purpose models switch on
These duties already existed. From this date the authorities can actually enforce them.
- 2 December 2026pending publication
Watermarking grace period ends for older generative systems
Generative AI systems already on the market before August 2026 must mark their output as machine-readable AI. Agreed by Parliament and Council, awaiting official publication.
- 2 December 2026pending publication
New ban on AI-generated non-consensual intimate imagery
A new prohibited practice is added to the Act. Agreed by Parliament and Council, awaiting official publication.
- 2 August 2027in force
Grandfathering deadline for existing general-purpose AI models
Providers of general-purpose models already on the market must be fully in line by this date. This deadline is already fixed in law and unaffected by the reform.
- 2 December 2027pending publication
Stand-alone high-risk systems (the eight high-risk categories) apply
The core rules for high-risk AI, such as hiring or credit-scoring tools sold on their own, apply from this date. Moved from the original 2 August 2026 date by the reform, and pending official publication.
- 2 August 2028pending publication
High-risk AI built into already-regulated products applies
High-risk AI embedded in products that already carry EU safety rules, like medical devices or machinery, applies from this date. Pending official publication.
The Digital Omnibus
Why some dates are moving, explained honestly
The EU concluded that the original deadlines for high-risk AI were too tight. The detailed technical standards that companies need in order to comply were not going to be ready in time, so it agreed to push those deadlines back through a reform package known as the Digital Omnibus.
The sequence, in plain terms: the European Commission proposed the change in November 2025. The Parliament and Council reached political agreement on 7 May 2026. The Parliament held its final vote on 16 June 2026, passing it by a large majority. The Council, as the final co-legislator, formally adopted it on 29 June 2026. That was a formal adoption, not a preliminary green light.
One step remains: signature and publication in the EU's Official Journal. That publication is what actually makes the new dates legally binding, and it needs to happen by roughly 30 July 2026 for the new dates to take effect before the original 2 August 2026 deadline arrives.
Status: not yet published
As of July 2026, the reform has not yet been signed and published in the Official Journal. Until it is, the original 2 August 2026 date for high-risk systems is still, technically, the one on the books.
Confirmed no matter what
The transparency duties are not delayed
The Digital Omnibus only touches the high-risk track: the eight high-risk categories and high-risk AI built into already-regulated products. It does not delay the transparency duties.
Chatbot disclosure and AI-content labelling still apply from 2 August 2026, whatever happens with the reform. If you run a customer-facing chatbot or publish AI-generated content, that date is fixed.
The one narrow exception: a machine-readable watermarking sub-duty for generative AI systems already on the market before August 2026 gets a grace period to 2 December 2026, and that single detail is part of the pending reform text.
Still pending
Agreed, but not yet law
These are politically agreed but not yet in force. They only become binding once the reform is signed and published in the Official Journal.
- pending publication
The high-risk delay dates (December 2027 and August 2028)
The move of stand-alone high-risk systems to 2 December 2027, and of high-risk AI embedded in already-regulated products to 2 August 2028.
- pending publication
The watermarking grace period (December 2026)
The extra time for generative AI systems already on the market before August 2026 to mark their output as machine-readable AI.
- pending publication
The new ban on AI-generated non-consensual intimate imagery (December 2026)
A new prohibited practice added to the Act.
Agreed by Parliament and Council, awaiting signature and Official Journal publication.
Not sure which of these dates apply to you?
The dates that matter depend on what your business actually does with AI. Take the free Exposure Check for a personal answer in about five minutes.
Frequently asked questions
Is the EU AI Act already in force, or is it still coming?
Both. The law entered into force in August 2024, and several parts are already live, including the banned practices, the AI literacy duty, and the rules for general-purpose AI models. Other parts, especially the high-risk rules, phase in later, and one of those dates is currently being pushed back.
What is the Digital Omnibus, and does it cancel the AI Act?
No. The Digital Omnibus is an EU reform package that adjusts some deadlines, mainly delaying the high-risk rules. It does not cancel the Act or the duties already in force. As of writing it has been agreed by Parliament and Council but not yet officially published.
When do I actually need to comply, given the delays?
The duties already in force apply now. Transparency duties for chatbots and AI-generated content apply from 2 August 2026, and that date is not affected by the delay. The high-risk rules are the part being moved, to 2 December 2027 for stand-alone systems, pending publication of the reform.
Why do these dates keep changing?
The main reason is that the technical standards companies need in order to comply with the high-risk rules were not going to be ready in time, so the EU agreed to push those deadlines back. The transparency and prohibition dates are not affected.
Where can I check the latest official date myself?
The authoritative source is the EU's Official Journal on EUR-Lex, where the reform will be published once signed. We update this page as the status changes, and note the date it was last current at the top.
Let's talk
Thirty minutes. Zero commitment.
Tell us what you need and we'll say exactly how we can help. If it isn't a fit, we'll save you the time.
This page is general information, not legal advice. For an answer specific to your business, take the Exposure Check or book a call.