Reports indicate that the National Security Agency is utilizing Mythos Preview, a model from Anthropic recently disclosed but not made publicly accessible, as per Axios. This development follows weeks after the NSA’s parent organization, the Department of Defense, identified Anthropic as a “supply chain risk” due to the company’s refusal to grant Pentagon officials full access to its model’s capabilities.

Earlier this month, Anthropic introduced Mythos as a cutting-edge model tailored for cybersecurity functions, claiming it was too powerful for public deployment due to its potential for conducting offensive cyberattacks. Consequently, the AI company restricted access to Mythos to around 40 entities, of which only a limited number have been publicly disclosed.

It appears the NSA is among those not publicly revealed, reportedly using Mythos mainly to identify exploitable vulnerabilities in environments. Additionally, the UK’s AI Security Institute has verified its access to Mythos.

The increasing utilization of Anthropic’s technologies by the U.S. military coincides with the argument made in court that these tools pose risks to national security. The disagreement with the Pentagon arose when Anthropic declined to enable Claude for widespread domestic surveillance and the development of autonomous weapons.

The NSA’s access to Mythos comes as Anthropic’s rapport with the Trump administration seems to be improving. Recently, Anthropic’s CEO Dario Amodei had a meeting with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent.

Reports indicate that the White House deemed the meeting successful.

TechCrunch has sought comments from the NSA. Anthropic has chosen not to respond.