NSA Reportedly Adopts Anthropic's Mythos Tool Despite Pentagon Security Concerns

The US National Security Agency (NSA) is using Anthropic's most advanced artificial intelligence model, Mythos Preview, even as the Department of Defense (DoD) officially labels the company a "supply chain risk," according to a report from Axios published over the weekend.

The disclosure underscores a growing split within the federal government, where military leadership has clashed with the AI firm, yet intelligence agencies appear to be deepening their reliance on its tools for cybersecurity work.

Axios reported that two sources said the NSA is using Mythos, while a third indicated the model has been adopted more broadly within the Department of Defense. Anthropic has restricted access to the model to roughly 40 organizations, citing the risk that its offensive cyber capabilities could be misused if widely distributed.

The company has publicly identified only 12 of those organizations. According to Axios, the NSA is among the unnamed agencies that received access, with most institutions reportedly using Mythos to scan for exploitable vulnerabilities in their environments. The United Kingdom's counterparts to the NSA have also confirmed access to the model through the country's AI Security Institute.

The tension dates back to February, when the DoD moved to cut off Anthropic and directed its vendors to do the same. The case is ongoing, and the Pentagon has argued in court that continued use of Anthropic's models could threaten US national security. The dispute flared during contract renegotiations earlier this year.

The Defense Department pressed Anthropic to make Claude available for what it called "all lawful purposes," while the company pushed to bar specific applications, particularly mass domestic surveillance and the development of autonomous weapons.

Some defense officials view Anthropic's stance as evidence that it cannot be trusted in critical military scenarios, a claim the company has rejected.

Anthropic chief executive Dario Amodei met with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent last Friday to discuss government deployment of Mythos and the company's broader security practices.

Both sides described the meeting as productive, according to Axios, with next steps expected to focus on how departments outside the Pentagon can engage with the model. Anthropic and the Pentagon declined to comment. The NSA and the Office of the Director of National Intelligence did not respond to inquiries.

The NSA's continued use of Mythos underscores the federal government's willingness to prioritize cybersecurity capabilities over procurement disputes. While the Pentagon has warned that Mythos could amplify cyberattacks due to its advanced coding and autonomous capabilities, other agencies appear to view the model as essential for hardening critical defenses.