EU AI Act Enforcement Powers Kick In for OpenAI, Anthropic, and Google

The European Commission has gained new powers under the EU AI Act to investigate general-purpose AI models, order changes, restrict EU market access, and fine providers, a milestone in the law’s staggered rollout that lands squarely on U.S. labs including OpenAI, Anthropic, and Google.

The authorities took effect on August 2, 2026. Regulators can now demand to evaluate models before public release in the region and fine a provider up to €15 million or 3% of annual turnover, whichever is higher.

Why the timing is tense

The move arrives amid already sharp U.S.-EU tech friction. In July, European regulators fined Google $1 billion under Digital Markets Act rules for preferential treatment of its own services, prompting U.S. President Donald Trump to threaten a “substantial” tariff response.

EU officials frame the AI powers as risk management at a new scale. Henna Virkkunen, executive vice-president for tech sovereignty, security and democracy, said harms can occur if AI is not properly designed and used, and that the most advanced models create risks “on an entirely new scale.”

Legal experts note that geography is not a shield. “A U.S. address does not put a lab outside the EU regulator’s reach,” said Elisabetta Righini, partner at Sidley Austin. Non-EU providers must appoint an EU-based authorised representative, and refusal to answer information requests or blocking a model evaluation can itself be fineable.

Labs already under the microscope

Europe has been pressing for access to Anthropic’s Mythos model for months. Reuters also reported the EU was in talks with OpenAI and Anthropic after cyber-related incidents involving their models; OpenAI confirmed it was in contact with the EU AI Office.

“We’ve collaborated closely with the European Commission and the wider ecosystem on implementing the AI Act, including its Codes of Practice, and will continue working together to help Europe realise the benefits of the Intelligence Age.” Tom Duff Gordon, VP, EMEA policy, OpenAI

Google said it remains dedicated to meeting applicable rules as part of advancing European AI infrastructure and innovation.

Decoded Take

Europe is no longer debating whether frontier AI should be regulated, it is building the enforcement machinery. Pre-release evaluation rights matter as much as the fine schedule: they give Brussels a lever before models reach consumers, not only after harm occurs.

For U.S. labs, the practical effect is a dual-track world. Washington is pushing voluntary, partly classified review programs; Brussels is standing up inspect-and-fine authority with explicit market-access teeth. Companies shipping global models will need compliance architectures that satisfy both, or risk choosing markets. Watch whether the first formal GPAI investigations land on model safety claims, cybersecurity misuse, or process failures such as incomplete disclosures.