The new EU Artificial Intelligence Act marks a turning point in technological regulation. Its implementation is not immediate; instead, it follows a timeline that will shape companies, administrations, and developers over the coming years. Knowing the key dates is essential to anticipate obligations, prepare for changes, and understand the true scope of this pioneering regulation. In this timeline, we have compiled the most relevant milestones.The Calendar of the New EU AI Act: The Ultimate Step-by-Step Implementation Guide.
Key dates timeline
2024
July 12, 2024
The law is officially published in the Official Journal of the European Union, representing the formal notification of its existence.
(Article 113)
August 1, 2024
The AI Act enters into force. This date marks its formal activation, although specific obligations will be implemented in a staggered manner.
(Article 113)
November 2, 2024
Member States must identify and make public the list of authorities responsible for protecting fundamental rights in the field of AI. They must also notify the Commission and other Member States.
(Article 77.2)
2025
February 2, 2025
Prohibitions on AI systems considered to pose “unacceptable risk” begin to apply, along with AI literacy obligations (for developers, providers, and operators).
(Article 113, point a) and Recital 179)
May 2, 2025
The Commission must have codes of practice ready to guide providers—especially of general-purpose AI (GPAI) models—on transparency obligations, security, etc.
(Article 56.9, Recital 179)
August 2, 2025
A critical date with multiple milestones:
-
Rules regarding notified bodies (Chapter III, Section 4), GPAI models (Chapter V), governance(Chapter VII), confidentiality (Article 78), and penalties (Articles 99 and 100) begin to apply.
-
Providers of GPAI models already on the market or in service before this date have until August 2, 2027, to adapt to the regulation.
-
Member States must designate national competent authorities (notification and market surveillance), inform the Commission, and publish their contact details.
-
They must also establish the national penalty regime (fines, enforcement measures) applicable under the Regulation.
-
If codes of practice are not approved or are deemed insufficient, the Commission may issue implementing acts to establish common rules.
-
From this point forward, the Commission will review prohibitions annually.
(Article 113, point b); for GPAI models: Article 111.3; for national authorities: Article 70; for penalties: Articles 99 and 100; for implementing acts: Article 56.9 and Recital 179; for annual review: Article 112.1)
At the time of publishing this article, we are here. I’d like to recommend our post where you can learn how AI helps anticipate your users’ needs: Predictive AI in Apps: The Future of User Experience
2026
February 2, 2026
The Commission must publish guidelines clarifying the practical application of Article 6, particularly regarding post-market monitoring.
(Article 6.5 and Article 72.3)
August 2, 2026
The majority of the remaining Act becomes active: unless explicitly excepted, most provisions of the AI Act become enforceable.
-
Operators of high-risk AI systems (not covered by GPAI models) must comply with the regulation whenever they make significant changes to their design from this date onward.
Each Member State must have at least one “regulatory sandbox”—a controlled environment for AI testing—operational to allow experimentation under regulatory supervision.
(Article 113 for general application; Article 111.2 for high risk; Article 57.1 for sandbox)
2027
August 2, 2027
Article 6(1) enters into force.
-
Providers of GPAI models marketed before August 2, 2025, must have complied with all obligations under the regulation.
AI systems integrated into large-scale IT infrastructures—as defined in Annex X—that were already in service before this date must comply with the regulation by December 31, 2030.
(Article 113 for Article 6(1); Article 111.3 for GPAI; Article 111.1 for large infrastructures)
2028 and beyond
August 2, 2028
The Commission will evaluate the performance of the new European AI authority (the “AI Office”).
From this date, every three years:
The impact and effectiveness of voluntary codes of conduct will be reviewed.
Every four years:
-
The Commission will analyze whether regulatory modifications are necessary (e.g., lists of systems subject to additional transparency requirements (Article 50), Annex III headings, or the governance model).
-
The delegation of powers conferred to the Commission will also be reviewed in this cycle, unless the Parliament or the Council objects.
August 2, 2029 (and every four years thereafter):
The Commission will submit a general evaluation report of the Regulation to the European institutions.
👉 Contact our team and let us advise you on creating the product your user needs.
Would you like me to adapt the internal link text to match a specific URL or help you draft the English version of the “Predictive AI” post as well?
