The National Security Agency (NSA) is reportedly using Anthropic’s Mythos Preview, a model recently announced but withheld from public release, according to Axios. This follows the Department of Defense labeling Anthropic as a “supply chain risk” for denying officials full access to its model’s capabilities.
Announced earlier this month, Mythos is designed for cybersecurity but deemed too capable of offensive cyberattacks for public release. Consequently, Anthropic limited access to 40 organizations, publicly naming only a dozen.
NSA is believed to be using Mythos primarily for scanning environments for vulnerabilities. The UK’s AI Security Institute also confirmed access to Mythos.
While the U.S. military’s use of Anthropic’s tools increases, it argues in court these tools could threaten national security. The dispute began when Anthropic declined to make its Claude model available for mass surveillance and autonomous weapons development.
Amid this, Anthropic’s relationship with the Trump administration seems to improve. Anthropic CEO Dario Amodei met with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent, with the meeting described as productive.
TechCrunch has contacted the NSA for comment, while Anthropic declined to respond.