AI Act: EU gains power to inspect and fine AI model providers

The European Commission can now inspect advanced AI models, limit their access to the EU market and impose significant fines on providers such as OpenAI, Anthropic and Google.

Kai Nicol-Schwarz reports for CNBC that the new supervisory powers for general-purpose AI models took effect on Sunday as part of the EU AI Act’s phased implementation.

The Commission may ask companies to submit models for evaluation before they are publicly released in the EU. It can also restrict a model’s market access where it identifies serious risks. Providers that breach the rules can face fines of up to €15 million or 3% of annual global turnover, whichever amount is higher.

The rules apply to any company offering a general-purpose AI model in the EU, regardless of its headquarters. Companies based outside the bloc must appoint an EU-based authorised representative to act as a regulatory contact point.

More than model safety

Henna Virkkunen, the Commission’s executive vice-president responsible for tech sovereignty, security and democracy, says increasingly capable AI models can create risks “on an entirely new scale” if they are poorly designed or used.

Legal experts say companies can also face penalties for failing to cooperate with regulators. Elisabetta Righini, a partner at Sidley Austin, told CNBC that refusing information requests, providing misleading responses or obstructing a model evaluation can each trigger fines.

The measures may increase friction between Brussels and Washington. The EU has stepped up its efforts to reduce reliance on US technology, while US officials have criticised European regulation and competition fines targeting American firms.

OpenAI says it has worked with the European Commission on implementing the AI Act and its codes of practice. Google says it remains committed to complying with applicable rules while expanding European AI infrastructure and innovation.