Anthropic’s Mythos model is reportedly in use by the NSA

The National Security Agency (NSA), the U.S. intelligence agency, is reportedly using Mythos Preview, a restricted AI model developed by Anthropic. According to Axios, the NSA is using the model despite an ongoing dispute with its parent agency, the Department of Defense (DoD). The DoD officially labeled Anthropic a “supply-chain risk”—an official designation used by the military department—after the company refused to allow Pentagon officials unrestricted access to the full capabilities of its models. This designation highlights the ongoing disputes over the company’s refusal to allow unrestricted access to its models.

Anthropic announced Mythos earlier this month as a frontier model—a term for a high-capability AI model—designed for cybersecurity tasks. However, the company withheld the model from public release, asserting that Mythos was “too capable of offensive cyberattacks” to be distributed openly. Consequently, Anthropic limited access to Mythos to around 40 organizations, of which it has publicly named only a dozen. The NSA is reportedly among the undisclosed recipients, using the model primarily for scanning environments for exploitable vulnerabilities. Meanwhile, the U.K.’s AI Security Institute has also confirmed it has access to Mythos.

The U.S. military continues to argue in court that Anthropic’s tools “threaten national security.” The dispute originated after Anthropic refused to make its models available for mass domestic surveillance and autonomous weapons development. Despite these ongoing legal tensions, Anthropic’s relationship with the Trump administration appears to be thawing. Last Friday, Anthropic chief executive Dario Amodei met with White House chief of staff Susie Wiles and Secretary of the Treasury Scott Bessent. The White House reportedly called the meeting productive, though Anthropic declined to comment and TechCrunch has reached out to the NSA for comment.

Why it matters

The situation highlights the friction between government agencies and AI labs, where national security demands for cybersecurity tools clash with broader regulatory concerns about supply-chain risks and model safety.