Too hot to handle: US government restricts next-gen AI model after troubling discovery

An AI model has reportedly exposed vulnerabilities in secure government systems as concerns over cybersecurity rage through top companies and government leadership alike.

The Associated Press reported that Mythos, an advanced Anthropic AI model, identified weaknesses in “sensitive and secure U.S. government computer systems during a testing exercise,” citing a U.S. official.

‘This tool broke into almost all of our classified systems, not in weeks but in hours.’

The official, who spoke on condition of anonymity, reportedly told the AP that the testing was conducted as a part of Project Glasswing, though the identification of the vulnerabilities does not mean that they were exploited.

Project Glasswing, which was announced on April 7, was described at the time as “an urgent attempt to put [the capabilities of frontier model Mythos Preview] to work for defensive purposes.”

RELATED: $965 billion AI giant warns we need to hit the brakes — but will China?

Andrew Harnik/Getty Images

The project announcement noted that Anthropic would be partnering with several industry giants to identify the vulnerabilities on their platforms, with Anthropic even pledging “up to $100M in usage credits for Mythos Preview across these efforts, as well as $4M in direct donations to open-source security organizations.”

The companies listed include Amazon Web Services, Anthropic, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks.

These defensive measures, including the testing performed by Mythos, was alluded to by Sen. Mark Warner (D-Va.) during a June 11 Senate Committee on Banking, Housing, and Urban Affairs hearing.

Mentioning Mythos and Anthropic by name, Warner said, “This tool broke into almost all of our classified systems, not in weeks but in hours.” Warner attributed this information to the head of the National Security Agency and U.S. Cyber Command, General Joshua Rudd.

The government limited access to Anthropic’s latest models for foreign nationals just days after President Trump signed an executive order establishing a national security framework for AI, cybersecurity, and the regulation and classification of “frontier models.” That directive resulted in the company disabling the model temporarily.

Blaze News reached out to the office of the National Cyber Director and Anthropic but did not immediately receive a response.

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!