The US Cybersecurity and Infrastructure Security Agency (CISA) is using Anthropic's powerful Mythos artificial intelligence model to scan government software for hidden weaknesses, three people familiar with the matter told Reuters, as per a recent report.

The scans are being run by CISA's Attack Surface Evaluation team, a unit that normally tests federal computer systems for security gaps that could be exploited by hackers or foreign spies. Two of the sources said the audits have already uncovered a large number of bugs, although they did not give details on how serious or widespread the problems were.

The deployment highlights growing government interest in Anthropic's tools, even though the company has had a tense relationship with Washington. Earlier this year, the Pentagon labelled Anthropic a supply-chain risk after it refused to let its AI be used for autonomous weapons or mass surveillance. That restriction was later blocked by a judge, and tensions have since eased.

The National Security Agency has also been testing Mythos, according to earlier reports, as per Reuters. Last week, Anthropic restored public access to a restricted version of the model after the White House lifted export controls it had imposed weeks earlier. CISA did not comment, and Anthropic did not respond to Reuters' questions.