AI Office

The EU AI Office became operational in February 2024 within the European Commission's Directorate-General for Communications Networks, Content and Technology. Its core mandate is to enforce the AI Act's rules on General Purpose AI models — the class that includes large language models such as those developed by OpenAI, Google, Anthropic, Mistral AI, and Aleph Alpha. In August 2025 the Office gained full enforcement powers, allowing it to investigate, audit, and fine providers of GPAI models that fail to meet transparency, risk assessment, and incident-reporting obligations.

The Office coordinates between member-state national supervisory authorities and serves as the lead enforcer for GPAI obligations at Union level. It maintains a registry of GPAI models, sets technical standards in conjunction with the European AI Office's advisory panels, and can request detailed documentation from model providers. Fines under the GPAI provisions can reach up to 3% of global annual turnover for violations, or 1% for failure to supply information.

The AI Office's creation reflects the EU's bet that central enforcement of foundational model rules is more effective than delegating to 27 separate national bodies. Critics — including European AI startups — argue that the compliance burden disproportionately affects smaller European providers such as Mistral AI and Aleph Alpha relative to the US hyperscalers whose compliance teams number in the hundreds.