The EU is now able to examine AI models, prohibit market entry, and impose fines on providers amounting to 3% of their revenue. The grace period has ended.

      TL;DR: The enforcement powers of the AI Act took effect on August 2. The Commission can require model evaluations, limit market access, and impose fines on providers up to 3% of their global revenue. This applies to companies like OpenAI, Anthropic, and Google, regardless of their headquarters' location.

      On August 2, the European Commission acquired enforcement authority over general-purpose AI models, marking the end of the grace period for Chapter V of the EU AI Act. The Commission can now mandate evaluations of models prior to their release in the region, restrict access to the EU market, and impose fines up to €15 million or 3% of the global annual revenue, whichever is greater. These powers affect any company providing general-purpose AI models in the EU, irrespective of their base of operations. Companies outside the EU must designate an EU-based authorized representative.

      According to Elisabetta Righini, a partner at Sidley Austin, "A US address does not exempt a lab from EU regulators' jurisdiction." She noted that liability can arise beyond significant violations: “Ignorance of an information request, providing misleading responses, or preventing a model evaluation can incur fines by themselves.” The Commission is currently in discussions with OpenAI and Anthropic following recent cyber incidents linked to their models, as reported by Reuters. The EU had requested access to Anthropic’s Mythos model for several months before the company consented to share it in June.

      This development heightens transatlantic tensions. Following a €1 billion fine imposed on Google under the Digital Markets Act in July, Trump threatened the EU with “substantial” tariffs. Incorporating AI Act enforcement into this situation provides Brussels with additional leverage over American tech firms. Anthropic’s hesitation in sharing open weights led to claims that it sought regulatory protection, and EU enforcement now equips the Commission with tools to demand the model access that Anthropic was hesitant to offer willingly.

      Tom Gordon, OpenAI's VP of EMEA policy, stated, “We’ve worked closely with the European Commission and the broader ecosystem to implement the AI Act.” Google expressed its commitment to complying with all relevant regulations. The White House is taking steps to control who accesses frontier models, putting American AI labs in a position where they must respond to two governments demanding pre-release reviews from both sides of the Atlantic, each with distinct requirements, timelines, and penalties. Companies that have created models for global use are learning that they must seek prior approval.