Definition · Plain-language
AI Act timeline
The AI Act timeline is the phased schedule by which the EU AI Act’s provisions become applicable, staggered over several years after the regulation entered into force.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
Why the Act applies in phases
The EU AI Act entered into force on 1 August 2024, twenty days after publication, but it deliberately does not apply all at once. Instead its obligations switch on in stages over the following years. Phasing serves two purposes: it lets providers, deployers and national authorities build the capabilities, documentation and supervisory structures the regulation demands, and it allows supporting materials — harmonised standards, guidance and codes of practice — to mature before the heaviest obligations bite. The result is a multi-year runway in which different parts of the regime become enforceable at different times rather than on a single switch-on date.
The key application dates
The earliest obligations land first. From February 2025 the prohibitions on unacceptable-risk AI practices apply, together with duties on AI literacy. From August 2025 the rules for general-purpose AI (GPAI) models apply, alongside elements of the governance and penalties architecture. The bulk of the regulation, including the core requirements for high-risk AI systems, becomes applicable from August 2026. Some provisions follow later, into 2027, notably for certain high-risk systems tied to products already regulated under other EU safety legislation. The exact obligations attaching to each milestone are set out in the regulation itself.
What the phasing means in practice
For organisations, the staggered timeline means compliance is not a single deadline but a sequence. The earliest priorities are confirming that no prohibited practices are in use and meeting AI-literacy duties, then preparing for GPAI obligations where relevant, and building toward the demanding high-risk requirements that arrive later. Because the dates are fixed in the regulation while supporting standards and guidance continue to develop, organisations commonly track both the legal milestones and the maturing technical materials. This overview describes the published schedule for general understanding and is not a substitute for legal advice on any particular obligation or date.
Key facts
At a glance
- Definition: the phased schedule of EU AI Act application dates
- In force: 1 August 2024 (entry into force)
- Feb 2025: prohibited AI practices and AI literacy duties apply
- Aug 2025: general-purpose AI (GPAI) obligations apply
- Aug 2026: most provisions, including core high-risk requirements, apply
- Into 2027: certain high-risk and product-linked provisions follow
Common misconceptions
What people often get wrong
Often heard: The whole EU AI Act applied the moment it entered into force.
Actually: Entry into force on 1 August 2024 started the clock, but obligations apply in phases — prohibited practices from February 2025, GPAI from August 2025, most high-risk rules from August 2026, and some provisions into 2027.
Often heard: High-risk AI requirements were among the first to apply.
Actually: The earliest obligations are the prohibitions and AI-literacy duties. The core high-risk requirements apply later, from August 2026, with certain product-linked high-risk provisions following into 2027.
Often heard: Because some dates are in the future, the Act can be ignored for now.
Actually: Several obligations, including the prohibitions and AI-literacy duties, already apply, and preparing for later high-risk requirements takes time. The phased schedule sequences compliance rather than postponing it entirely.
Going deeper
