The European Union aims to “intensify” talks with the US government on cutting-edge AI models, including those with “cyber capabilities,” a European Commission official told CNBC.
Anthropic’s powerful Mythos model, announced in April, infuriated governments and businesses and sparked a wave of concerns about AI-powered cyberattacks. That same month, the Wall Street Journal reported that the White House opposed Anthropic’s plan to expand access to its powerful Mythos model.
Anthropic said Thursday it plans to offer a “Mythos-class” model to customers in the coming weeks.
The White House said it is working closely with the AI Institute to balance innovation and safety as the United States seeks to maintain its lead over China in the global AI race.
Anthropic initially rolled out the model in preview to a select group of companies and organizations as part of an initiative called Project Glasswing. However, the AI Institute has not yet granted preview access to review it to any government agency outside the EU, its AI agency, or the United States, with the exception of the UK’s AI Security Institute.
Anthropic is proposing to allow about 70 additional companies and organizations to use Mythos, a move that government officials have opposed, citing security concerns, the WSJ reported.
“Cybersecurity is a common priority and we have agreed to mutually recognize our respective standards in this area,” European Commission spokesman Thomas Renier told CNBC, referring to discussions between the Commission and the U.S. government.
“In parallel, we are currently expanding technical discussions with Anthropic and other model developers who are communicating their latest models to the AI Office.”
A person familiar with the talks between the EU and Anthropic told CNBC that during discussions about preview access to Mythos, Anthropic told the commission that the European Union would first need to seek permission from the U.S. government. The official requested anonymity because he is not authorized to discuss confidential discussions publicly.
The official added that Mr. Anthropic told the committee that while the White House is not specifically opposed to sharing Mythos with the European Union, the administration is generally opposed to companies sharing Mythos with governments other than the United States.
When asked for comment Thursday, the White House Press Office referred CNBC to Treasury Secretary Scott Bessent’s briefing comments. Bessent said the U.S. government is working “very closely” with the AI Institute.
“They’re a great partner and we’re going to have a solution that solves calculus to the fullest,” Bessent said, adding that the United States is the world’s AI leader, with China in second place. “We want to make sure we maintain this lead. That’s why we’re working on a precise calculation between innovation and safety, and we want to optimize it.”
Models of Mythos functionality require strong cyber protection before being released to the public, Anthropic said Thursday. “We are moving quickly to develop these safety measures and expect to have Mythos-class models available to all customers in the coming weeks.”
CEO Dario Amodei warned earlier this month that the company has six to 12 months to patch tens of thousands of software vulnerabilities discovered in its Mythos model before China’s AI catches up.
Antropic declined to comment.
