EU AI Act Timeline: Which Rules Apply and When?
The EU AI Act applies in stages, beginning with prohibited practices and AI literacy in 2025 and moving through general-purpose AI and broader system obligations.
Timeline
- February 2, 2025: Prohibited AI practices and AI-literacy provisions started applying.
- August 2, 2025: Governance rules and obligations for general-purpose AI models began applying.
- August 2, 2026: Most remaining provisions became applicable, with some later exceptions in the staged timetable.
The EU AI Act did not begin applying all at once. Its obligations phase in by date and by the kind of AI involved. The European Commission’s implementation timeline is the safest starting point because a short summary can easily mix obligations for providers, deployers, general-purpose models and high-risk systems. [1]
On February 2, 2025, rules on prohibited AI practices began applying, along with the requirement for providers and deployers to take measures supporting an appropriate level of AI literacy among relevant staff. This date did not make every AI system subject to every operational requirement in the regulation. [1][2]
On August 2, 2025, provisions concerning governance and general-purpose AI models began applying. The Act distinguishes general-purpose models from systems built for a specific use. Additional duties can apply to general-purpose models with systemic risk, so a company’s role and the model’s classification matter. [1]
Most other provisions became applicable on August 2, 2026, including a broad set of transparency and enforcement milestones. Certain high-risk system obligations follow later dates in the official schedule. A product can also be affected by separate EU product-safety, privacy, consumer or sector rules. [1][2]
“Applies” is different from “entered into force.” The law entered into force in 2024, while compliance dates were deliberately staggered. Organizations should map each use case, identify whether they are a provider, importer, distributor or deployer, and document the basis for the classification. [1][2]
This guide is a general timeline rather than legal advice. The Commission’s AI Act Service Desk and the regulation itself should be checked for the exact article, transition rule and any implementing guidance relevant to a particular system. [1][2]
Companies outside the European Union can still be affected when they place an AI system or general-purpose model on the EU market, or when the output is used in the Union under the regulation’s scope. Territorial coverage and contractual position require fact-specific analysis; simply hosting a model outside Europe does not by itself answer whether the Act applies. [1][2]
Sources
- European Commission — EU AI Act implementation timeline
- European Commission — When enforcement starts