Putting Europe’s AI law into practice
Technology & Science
Last updated
The big picture
The European Union’s AI Act sets rules for artificial intelligence according to the risks associated with its use. It addresses issues such as prohibited practices, higher-risk applications, general-purpose models and transparency when people encounter AI-generated material. Those are different categories, so there is no single compliance date that explains the whole law. Implementation also depends on technical standards, guidance and national authorities capable of enforcing the rules. Delays in those preparations helped lead to changes in the timetable through the 2026 AI Omnibus. For readers, the key question is which obligation applies to which system, rather than whether AI has become either completely regulated or completely unregulated on one particular day.
Where things stand
Many provisions now apply, but the Commission’s updated timetable places high-risk-system requirements on December 2, 2027, and requirements for AI embedded in regulated products on August 2, 2028. The July 2026 amendments changed earlier deadlines. Older explainers that put all high-risk obligations in August 2026 can therefore be misleading.
Key facts
- The rules distinguish AI providers from organizations deploying systems in practice.
- Transparency duties and high-risk-system duties follow different provisions and dates.
- The Council’s updated timeline also identifies a December 2026 transparency transition deadline.
How we got here
Prohibitions and AI-literacy provisions begin applying.
Governance and general-purpose-model obligations begin applying.
The AI Omnibus enters into force, changing parts of the implementation timetable.
Further AI Act provisions begin applying, subject to the amended exceptions.
What happens next
- The updated high-risk deadlines are December 2, 2027, and August 2, 2028, depending on the category of system.
Related Daily Edition coverage
- No related edition articles have been linked yet.