The EU can now inspect AI models, block market access, and fine providers 3% of turnover. The grace period is over.

The European Commission gained enforcement powers over general-purpose AI models on August 2, ending the grace period for Chapter V of the EU AI Act. The Commission can now demand to evaluate models before regional release, restrict EU market access, and fine providers up to €15 million or 3% of global annual turnover, whichever is higher. The powers apply to any company offering a general-purpose AI model in the EU, regardless of where it is headquartered. Non-EU providers must appoint an EU-based authorised representative.

“A US address does not put a lab outside the EU regulator’s reach,” said Elisabetta Righini, partner at Sidley Austin. She added that liability extends beyond substantive breaches: “Refusing an information request, giving misleading answers, or blocking a model evaluation is finable on its own.” The Commission is in talks with OpenAI and Anthropic after recent cyber incidents involving their models, Reuters reported. The EU had sought access to Anthropic’s Mythos model for months before the company agreed to share it in June.

The timing raises transatlantic tensions. Trump threatened the EU with “substantial” tariffs after Brussels fined Google $1 billion in July under Digital Markets Act rules. Adding AI Act enforcement to that dynamic gives Brussels another lever against American tech companies. Anthropic’s silence on open weights drew accusations that it wanted regulatory protection, and EU enforcement gives the Commission tools to demand the kind of model access Anthropic was reluctant to provide voluntarily.

“We’ve collaborated closely with the European Commission and the wider ecosystem on implementing the AI Act,” said Tom Gordon, VP of EMEA policy at OpenAI. Google said it remains “dedicated to meeting all applicable rules.” The White House is also asserting control over who accesses frontier models, and American AI labs now face two governments demanding pre-release review from opposite sides of the Atlantic, each with its own thresholds, timelines, and penalties. The companies that built models to serve the world are discovering that the world wants to approve them first.