The EU AI Act entered into force on 1 August 2024 and applies through a phased timetable in Article 113, as amended by Regulation (EU) 2026/1744. General provisions including AI literacy and the original prohibited practices apply from 2 February 2025, general-purpose AI model rules and governance provisions from 2 August 2025, the majority of rules including Article 50 transparency from 2 August 2026, two new prohibitions from 2 December 2026, Annex III high-risk rules from 2 December 2027, and Annex I high-risk rules from 2 August 2028.
What is the timeline at a glance?
The timeline is phased: entry into force, early application, general-purpose model and governance application, general application, a December 2026 stage, and two later high-risk stages. The Act entered into force on 1 August 2024, which means it became law on that date, but most obligations did not begin to apply until the dates Article 113 sets for each provision. Regulation (EU) 2026/1744, the Digital Omnibus on AI, signed on 8 July 2026, published in the Official Journal on 24 July 2026, and in force from 27 July 2026, amended those application dates, deferring the high-risk stages and adding the December 2026 stage. Every date in this guide is stated with its provision, and the consolidated text should be checked, because pre-amendment summaries give different high-risk dates and omit 2 December 2026.
What happened on 1 August 2024?
On 1 August 2024, Regulation (EU) 2024/1689 entered into force, twenty days after its publication in the Official Journal. Entry into force is the date the regulation became part of Union law. It is not the date on which every obligation began to apply. From that date, organisations could begin mapping systems, roles and classifications against the final text, and the institutions could begin the governance and standardisation work the Act envisages. Businesses that treated 1 August 2024 as a compliance deadline for every provision misunderstood the phased structure, and businesses that treated it as irrelevant until 2026 missed the preparation time the phased structure was designed to provide. The distinction between entry into force and application is the key to reading every later date correctly.
For the broader regulatory framework, see EU AI Act timeline.
What began to apply on 2 February 2025?
On 2 February 2025, the general provisions in Chapter I, including the definitions in Article 3 and AI literacy in Article 4, and the original prohibited practices in Article 5(1)(a) to (h), began to apply. AI literacy requires providers and deployers to take measures to ensure a sufficient level of AI literacy for staff and others dealing with the operation and use of AI systems, taking account of technical knowledge, experience, education and training and the context of use. The original prohibitions, covering manipulative techniques, exploitation of vulnerability, social scoring, predictive policing, facial image scraping, emotion inference, biometric categorisation and real-time remote biometric identification, became applicable on this date. This was the first significant operational milestone, and it arrived six months after entry into force.
What began to apply on 2 August 2025?
On 2 August 2025, the general-purpose AI model rules in Chapter V began to apply, and the governance framework had to be in place, including the AI Office, the European Artificial Intelligence Board, the advisory forum, the scientific panel, and the designation of national competent authorities. Providers of general-purpose AI models faced the Chapter V duties on technical documentation, information for downstream providers, copyright policy and training content summary, with additional duties for models with systemic risk. Transitional provisions in Article 111 applied to general-purpose AI models already placed on the market before 2 August 2025, in the circumstances that article sets. The date marked the point at which model providers, as distinct from system providers, entered an active compliance phase.
What began to apply on 2 August 2026?
On 2 August 2026, the majority of the remaining rules began to apply, including the Article 50 transparency obligations for providers and deployers of certain AI systems, and enforcement began for the applicable rules, including prohibitions, general-purpose model duties, transparency and AI literacy. Regulation (EU) 2026/1744 entered into force shortly before, on 27 July 2026, and amended the later high-risk dates, but it did not move the 2 August 2026 general application stage. The stage is sometimes described as the date the Act came into force, which is inaccurate: the Act entered into force in 2024, and 2 August 2026 is a major application stage. Organisations should treat the duties that apply from this date as already applicable, and should not describe them as future obligations.
Continue with what the EU AI Act is.
What applies from 2 December 2026 to 2 August 2028?
On 2 December 2026, the two new Article 5 prohibitions in points (ba) and (bb), added by Regulation (EU) 2026/1744, begin to apply, and the transition period ends for certain providers of AI systems generating synthetic content already placed on the market before 2 August 2026 to comply with the Article 50(2) marking and detection obligation. Member States should have at least one AI regulatory sandbox operational by 2 August 2027, in the Service Desk wording, which states should have rather than imposing a hard obligation in those terms. Chapter III Sections 1 to 3 high-risk rules apply for Annex III systems under Article 6(2) from 2 December 2027, and for Annex I product-embedded systems under Article 6(1) from 2 August 2028. The December 2027 and August 2028 dates were 2 August 2026 and 2 August 2027 before the Omnibus amendments, and the change should be reflected in every planning document.
How should companies use this timeline, and what does it not decide?
Companies should use the timeline as a planning input, not as a compliance conclusion. For each system, they should record the provision, its application date, any transitional provision, the official source, and the date last checked, and should assign an owner and a review date. Inventory and classification should be done first, because they take time and because earlier duties already apply, even where a particular high-risk requirement applies later. A staged plan should prioritise prohibited practices, AI literacy, general-purpose model duties and transparency duties that already apply, and should plan evidence and handoffs for the later high-risk milestones. This guide states dates from the verified list and the consolidated text. It does not decide the application date for any particular system, which depends on its classification, its placement date, and any transitional provision, confirmed with a competent authority or legal adviser.
Frequently asked questions
Did the entire EU AI Act start applying in 2024?
No. The Act entered into force on 1 August 2024 but uses a phased application schedule in Article 113, as amended by Regulation (EU) 2026/1744, with different provisions becoming applicable at different times through 2028.
Was 2 August 2026 the final AI Act deadline?
No. Although 2 August 2026 is a major application stage, later dates remain relevant, including 2 December 2026 for new prohibitions and the Article 50(2) transition, 2 December 2027 for Annex III high-risk rules, and 2 August 2028 for Annex I high-risk rules.
Why should companies prepare before their AI Act deadline?
Activities such as system inventory, role determination, risk classification, documentation and governance design can require substantial preparation before the relevant requirements begin applying, and earlier duties in the phased timetable already apply to many systems.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689 ↗AI Act Service Desk timeline ↗European Commission regulatory framework for AI ↗