The EU AI Act Enters Into Force Across the European Union

How the risk categories work

Most AI systems, such as spam filters and recommender systems, fall into a minimal risk category with no obligations. Systems with specific transparency risk, such as chatbots, must disclose that users are interacting with a machine, and certain AI-generated content, including deep fakes, must be labelled.

High-risk systems, such as AI used in recruitment or to assess eligibility for a loan, must meet strict requirements including risk mitigation, high quality data sets, logging, detailed documentation, human oversight, and robustness, accuracy and cybersecurity. AI systems considered a clear threat to fundamental rights are banned, including social scoring and certain uses of biometric systems such as emotion recognition in the workplace.

The Act also sets rules for general-purpose AI models, aimed at transparency along the value chain and at the systemic risks posed by the most capable models.

Enforcement and fines

Member States have until 2 August 2025 to designate national competent authorities. The Commission's AI Office will be the key implementation body at EU level and will enforce the rules on general-purpose AI models. Fines can reach 7% of global annual turnover for violations involving banned applications, 3% for other obligations and 1.5% for supplying incorrect information.

What applies when

Most rules will apply from 2 August 2026. Prohibitions on AI systems presenting an unacceptable risk apply after six months, and the rules for general-purpose AI models after 12 months. Until then, the Commission's AI Pact invites developers to adopt key obligations voluntarily, and work has begun on a general-purpose AI Code of Practice.

The legislation follows the political agreement reached in December 2023 between the European Parliament and the Council. Businesses that build or deploy AI in the EU now have a fixed timetable for compliance.

Sources

• European Commission: European Artificial Intelligence Act comes into force
• European Commission: Commission welcomes political agreement on Artificial Intelligence Act