Français

EU AI Act timeline: what applies, and from when

Every deadline in the EU AI Act that binds an organisation, from February 2025 to December 2030, each with the legal basis behind it and the traps in reading them.

The EU AI Act does not apply all at once. Regulation (EU) 2024/1689 applies in stages set by the third paragraph of Article 113, and that schedule was rewritten on 27 July 2026 by Regulation (EU) 2026/1744.

This page gives every deadline that binds an organisation, with the provision it rests on. Which obligations attach to each is covered on the pages for each topic.

What is the EU AI Act timeline?

The Regulation applies in stages since 2 February 2025, with 2 August 2026 as its general date. Prohibitions and AI literacy have applied since February 2025, general-purpose AI models and the penalty regime since August 2025, Article 101 excepted, Article 50 transparency since 2 August 2026. High risk was postponed to December 2027 and August 2028.

Which date rests on which article?

Every deadline below carries its legal basis, because that is what lets you check a date rather than believe it. Two things are worth separating before reading any of it. Entry into force happened once, twenty days after publication in the Official Journal. Application is staged, and it is the staging that matters in practice.

DateWhat starts applyingBasis
2 February 2025Chapters I and II: AI literacy (Article 4), and the prohibitions in Article 5(1), points (a) and (b) and (c) to (h), with the full authorisation regime for real-time remote biometric identificationArt. 113, 3rd para., (a)
2 August 2025Chapter III Section 4, Chapter V (general-purpose AI models), Chapter VII (governance), Chapter XII (penalties) and Article 78, with Article 101 exceptedArt. 113, 3rd para., (b)
27 July 2026Articles 102 to 110, the amendments to sectoral Union legislationArt. 113, 3rd para., (d)
2 August 2026The general date. Everything not otherwise listed, including Article 50 and Article 101Art. 113, 2nd para.
2 December 2026The two prohibitions added by the omnibus (Article 5(1), points (ba) and (bb)) and the filter in Article 5(1a) and (1b)Art. 113, 3rd para., (a), as rewritten
2 December 2026Separately: providers of systems generating synthetic content placed on the market before 2 August 2026 must comply with Article 50(2) in fullArt. 111(4)
2 August 2027Providers of general-purpose AI models placed on the market before 2 August 2025 must complyArt. 111(3)
2 December 2027Chapter III, Sections 1, 2 and 3 for systems classified as high-risk under Article 6(2) and Annex IIIArt. 113, 3rd para., (c)(i)
2 August 2028The same, for systems classified as high-risk under Article 6(1) and Annex IArt. 113, 3rd para., (c)(ii)
2 August 2030Providers and deployers of high-risk systems intended to be used by public authorities must comply, in any eventArt. 111(2), 2nd sentence
31 December 2030Systems that are components of the large-scale IT systems established by the legal acts listed in Annex X, placed on the market or put into service before 2 August 2027Art. 111(1)

What are the traps in reading this calendar?

Four, and they all have the same shape: a date read without its article, or an article read without its exceptions. Article 50 has no deadline of its own; 2 December 2026 carries two distinct things; 2 August 2025 does not cover Article 101; and Article 6(5) is excluded from the postponement. Each is a mistake found in circulation, and each is settled by the text.

Each of these is a mistake found in circulation, and each is settled by the text.

Article 50 has no deadline of its own

It sits in Chapter IV, and Chapter IV appears nowhere in the list of exceptions in the third paragraph of Article 113. It therefore falls under the general date, 2 August 2026. Saying “Article 50 has applied since 2 August 2026” is proved by the absence of Chapter IV from that list, not by any sentence of the text that says so. The omnibus touched only paragraph 7 of Article 50, on codes of practice.

2 December 2026 carries two different things

On one side, two new prohibitions in Article 5, with their filter. On the other, a catch-up limited to paragraph 2 of Article 50, for systems placed on the market before 2 August 2026 only.

Different legal basis, different scope, different addressees. Presenting them as one deadline is the most common error on this date. The catch-up binds providers only, and only for systems generating synthetic content, and it covers Article 50(2) in full, that is marking and detectability, not marking alone.

2 August 2025 does not cover Article 101

Point (b) excepts it in terms: Chapter XII applies from 2 August 2025 “with the exception of Article 101”. Article 101 carries the fines the Commission may impose on providers of general-purpose AI models. Those therefore fall under the general date of 2 August 2026.

A related point of timing: an obligation can be enforceable before the penalty attached to it is. The prohibitions have applied since 2 February 2025, the penalty regime of Article 99 since 2 August 2025, six months apart.

Article 6(5) is excluded from the postponement

Point (c) postpones Chapter III, Sections 1, 2 and 3 “with the exception of Article 6(5)”. That paragraph requires the Commission, not businesses, to provide guidelines on the practical implementation of Article 6, together with “a comprehensive list of practical examples of use cases of AI systems that are high-risk and not high-risk”, no later than 2 February 2026.

The text does not say why that paragraph was carved out of the postponement, and we do not attribute a motive to the legislature: the established fact is the exclusion itself. What is also established is that the deadline has passed. Only a draft exists, put online on 19 May 2026: as at 24 August 2026, the date of our last check, no adopted version of the guidelines on high-risk classification was found, only a draft exists.

Do systems already in service have to meet these dates?

Usually not, as long as they undergo no significant design changes. Article 111(2) exempts systems placed on the market or put into service before the date of application, with no time limit — except for systems intended for use by public authorities, which must comply by 2 August 2030 in any event. And that grandfathering never covers Article 5.

The table above gives dates of application. Article 111 says to whom they apply where the system was already on the market, and without it, a reader would conclude that every high-risk system must comply by 2 December 2027 or 2 August 2028. That is wrong for most of the installed base.

“Without prejudice to the application of Article 5 as referred to in Article 113, third paragraph, point (a), this Regulation shall apply to operators of high-risk AI systems, other than the systems referred to in paragraph 1 of this Article, that have been placed on the market or put into service before the date of application of Chapter III referred to in Article 113, only if, as from that date, those systems are subject to significant changes in their designs. In any case, the providers and deployers of high-risk AI systems intended to be used by public authorities shall take the necessary steps to comply with the requirements and obligations laid down in this Regulation by 2 August 2030.”

Source: Article 111(2), consolidated version of 27 July 2026, as rewritten by Regulation (EU) 2026/1744

Three rules, which should not be run together.

A system already on the market is not caught until it changes. Placed on the market or put into service before the date of application of Chapter III, it falls under the Regulation only if, from that date, it undergoes “significant changes in their designs”.

Unless it is intended for public authorities. There the exemption is not open-ended: providers and deployers must have complied by 2 August 2030, whether or not any significant change occurred. This is not extra time added to the general dates; it is the boundary of an exemption.

And Article 5 is never covered. Paragraphs 1 and 2 of Article 111 both open with that reservation. A legacy system remains subject to the prohibitions from 2 February 2025, and from 2 December 2026 for the two points added by the omnibus. No transitional regime makes an exception to that.

What did 27 July 2026 change?

Regulation (EU) 2026/1744 entered into force on 27 July 2026. Two things follow that are easy to miss in a calendar of application dates.

Articles 102 to 110, which amend other Union legislation, apply from that same date under point (d).

And the postponements themselves took effect then: reading the July 2024 version of Article 113 today produces dates that are no longer the law.

What this timeline rests on

Articles 111, 112 and 113 of Regulation (EU) 2024/1689 in their consolidated version of 27 July 2026, as amended by Regulation (EU) 2026/1744, the authentic text remaining the one published in the Official Journal, the consolidated version having documentary value only. Quotations are reproduced from the official English text and are never translated from another language version. Every date on this page carries the provision it rests on; a date without a basis is a date not to be trusted.

Sources. Regulation (EU) 2024/1689, consolidated text of 27 July 2026 and Regulation (EU) 2026/1744, both consulted on 27 August 2026.

Frequently asked questions

When did the EU AI Act enter into force, and when does it apply?

Entry into force and application are two different things. The Regulation entered into force on the twentieth day following its publication in the Official Journal, in August 2024. Application is staged by the third paragraph of Article 113: 2 February 2025 for Chapters I and II, 2 August 2025 for the general-purpose AI model chapter and for the penalty regime, Article 101 excepted, 2 August 2026 as the general date, and 2 December 2027 and 2 August 2028 for high-risk systems.

What happens on 2 December 2026?

Two different things, and they should not be merged. First, the two prohibitions added by Regulation (EU) 2026/1744 (non-consensual intimate material, and child sexual abuse material or performance) together with the filter in Article 5(1a) and (1b), start to apply. Second, and on a separate legal basis, providers of systems generating synthetic content placed on the market before 2 August 2026 must comply with Article 50(2) in full. Different basis, different scope, different addressees.

Does the timeline apply to systems already on the market?

Not in the same way, and this is the most consequential thing the calendar alone does not tell you. Article 111(2) provides that the Regulation applies to operators of high-risk systems placed on the market or put into service before the date of application of Chapter III only if, as from that date, those systems are subject to significant changes in their designs. Two limits: systems intended to be used by public authorities must comply by 2 August 2030 in any event, and Article 5 is never covered by any transitional regime.

Read next

This content is a technical and regulatory information resource. It does not constitute legal advice.