The EU said Thursday it is in discussions with U.S. AI firm Anthropic over concerns about the capabilities of its latest model, which the company itself worries could be a boon for hackers.

Anthropic's new model, Claude Mythos, has proven keenly adept at exposing software weaknesses, pushing the company to postpone full release.

"We have a new AI model that is being released. It comes with a certain number of risks. We need information when it comes to these risks," European Commission spokesman Thomas Regnier told reporters.

"We're reaching out to the platform, to Anthropic. We have received certain information," Regnier added, confirming that a first meeting took place Wednesday and more would follow.

Anthropic said earlier this month it restricted the release of Mythos to just 40 major tech players to give firms a head start in fixing vulnerabilities before they could be exploited by attackers.

But no foreign entities were included, raising concerns about the world's preparedness for a model whose offensive capabilities would not stop at U.S. borders.

According to Anthropic and partners, Mythos can autonomously scan vast amounts of code to find and chain together previously unknown security vulnerabilities in all kinds of software, from operating systems to web browsers.

Crucially, they warn, this can be done at a speed and scale no human could match, meaning it could be used to bring down banks, hospitals or national infrastructure within hours.

Anthropic's announcement has encountered a mix of alarm and scepticism, with some pointing out the firm, one of several contenders in a fierce artificial intelligence race, stood to gain from the hype.

But the heads of America's biggest banks reportedly met this month with Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bessent to weigh the security implications of the yet-to-be released model.