European Commission launches the European AI Office

The European Commission formally established the European AI Office, a new body sitting inside the Commission and tasked with implementing and enforcing the EU’s AI Act once it took effect. The Office had been announced via a Commission decision the previous month and became operational as that decision entered into force.

Its remit centred on general-purpose AI models — the category the Act singled out for direct EU-level oversight rather than leaving to national regulators. The Office was given powers to investigate suspected violations, request technical documentation and training-data summaries from developers, conduct model evaluations, and, in cases of non-compliance, restrict a model’s availability in the EU market or issue fines. It was also charged with developing codes of practice and evaluation methodologies to give substance to the Act’s more general obligations, and with supporting the AI Act’s national-level governance bodies in member states.

Staffed with more than 100 people across units covering regulation, safety, innovation and sector-specific applications such as health, the Office represented the EU’s institutional answer to a problem other jurisdictions were addressing differently: rather than a network of national AI regulators, or the UK and US’s model of voluntary safety-institute evaluation without direct enforcement power, the EU built a centralised office with statutory authority over the largest models sold into its market.

Its practical significance depended on the AI Act’s own staggered rollout, with obligations for general-purpose AI models phasing in through 2025 and full enforcement powers arriving later still — meaning the Office’s creation marked the start of the EU’s enforcement apparatus rather than its first exercise of it.