NSA Adopts Anthropic's Mythos AI Despite Pentagon Supply Chain Ban

The National Security Agency (NSA) is leveraging Anthropic's Mythos Preview, the company’s most powerful AI model to date, despite objections from top officials at the Department of Defense (DoD). The DoD, which oversees the NSA, has labeled Anthropic a "supply chain risk," according to two sources who spoke with Axios.

This development underscores a critical tension: the government’s pressing cybersecurity demands appear to be overriding the Pentagon’s ongoing feud with Anthropic over supply chain security.

Pentagon’s Blacklist and Ongoing Legal Battle

In February, the Department of Defense moved to cut off Anthropic and mandated that its vendors follow suit. This directive remains in effect, with the military simultaneously expanding its use of Anthropic’s tools while arguing in court that these tools pose a threat to U.S. national security.

Scope of Mythos Usage Across Agencies

Two sources confirmed that the NSA is actively using Mythos, while a third indicated that the model is being adopted more broadly within the department. The specific applications of Mythos within the NSA remain unclear, though other organizations with access to the model primarily utilize it to scan environments for exploitable security vulnerabilities.

Anthropic has restricted access to Mythos to approximately 40 organizations, citing concerns that its offensive cyber capabilities could be misused if widely distributed. To date, the company has publicly disclosed only 12 of these organizations. One source confirmed that the NSA is among the unnamed agencies granted access.

Meanwhile, the U.K.’s AI Security Institute has also been granted access to Mythos, according to statements from British officials.

White House Engagement and Future Steps

On Friday, Anthropic CEO Dario Amodei met with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent to discuss the integration of Mythos within government operations and Anthropic’s broader security practices. Sources familiar with the meeting described it as productive, with next steps expected to focus on how departments beyond the Pentagon engage with the model.

Anthropic and the Pentagon declined to comment on the discussions. The NSA and the Office of the Director of National Intelligence also did not respond to requests for comment.

Root of the Conflict: Contract Renegotiations and Core Disputes

The rift between the Pentagon and Anthropic emerged during tense contract renegotiations earlier this year. The DoD demanded that Anthropic make its Claude model available for "all lawful purposes," while Anthropic insisted on restrictions to prevent mass domestic surveillance and the development of autonomous weapons.

Some Defense officials continue to argue that Anthropic’s stance demonstrates an inability to meet military needs when required. Anthropic has denied these claims, asserting its commitment to responsible AI development.

Within the administration, there is a growing sentiment that the dispute must be resolved to allow the utilization of Anthropic’s cutting-edge tools.