• 3 min read
EU gains power to inspect AI models before launch
The EU can inspect OpenAI, Anthropic, and Google models before launch, restrict market access, and impose fines of up to €15 million or 3% of turnover.
Image: iXBT
The European Commission can now demand pre-launch inspections of general-purpose AI models from OpenAI, Anthropic, and Google, restrict their access to the EU market, and impose fines of up to €15 million or 3% of a company’s annual global turnover, whichever is higher.
The powers arrive as the EU’s AI Act takes effect in stages. They apply to all providers of general-purpose AI models operating in the European Union, regardless of where those companies are incorporated. US headquarters do not put a provider outside the law’s reach, according to Elisabetta Rigini, a partner at law firm Sidley Austin. Foreign developers must also appoint an official representative inside the EU.
What the AI Act powers cover
The Commission may request access to a model before it launches in the European market. It can also limit market access if a provider fails to meet the law’s requirements.
Penalties are not limited to direct violations of the AI Act. The same enforcement regime can apply if a company refuses to provide information, misleads regulators, or obstructs an inspection. The potential maximum—3% of global annual turnover—could be substantially larger than the fixed €15 million figure for the biggest technology companies.
The new authority comes as the EU seeks greater technological independence and increases scrutiny of major US AI providers. Reuters, as cited by iXBT, reported that the EU is already negotiating with OpenAI and Anthropic following recent cyber incidents involving their models. OpenAI confirmed that it is in contact with the European AI Office.
Recommended reading
DeepMind says AI spending bets on self-improving machines
OpenAI, Anthropic, and Google all currently offer their models in both the United States and the European Union. Their representative consumer plans include ChatGPT Plus at $20 per month, Claude Pro at $20 per month when billed monthly—or $17 per month with annual billing—and Google’s Gemini Advanced offering at $19.99 per month in the US. Anthropic lists Claude Pro in the EU at €18 plus VAT per month.
That availability means the new inspection powers affect products already sold to European consumers, not merely future research systems. However, the research reviewed for this report did not establish whether EU checks will change public pricing or regional availability notices. OpenAI and Google also did not show a single explicit euro list price on the pages reviewed, leaving country-specific pricing unresolved.
Rising US-EU technology tensions
The regulatory move follows a broader escalation between Brussels and Washington. In July, European authorities fined Google nearly $1 billion under the Digital Markets Act. US President Donald Trump then threatened the EU with “substantial tariffs.”
Representatives of OpenAI and Google said they would continue cooperating with the European Commission and comply with the new legislation. For AI providers, that cooperation is no longer simply a policy preference: refusing information or blocking an inspection could itself trigger enforcement, while market access can now be made conditional on regulatory review.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via iXBT
