THE US National Security Agency is reportedly using Anthropic’s Claude Mythos, despite the Department of Defense viewing the vendor as a supply-chain risk. According to Axios, the NSA is already using Mythos Preview even as DoD officials push to cut ties with Anthropic, highlighting a tension between defence needs and procurement caution.
The article notes that Mythos is considered highly capable for cybersecurity, with access limited due to concerns it could be misused, but also valued for finding vulnerabilities and informing defensive measures. It describes Project Glasswing as a joint effort involving major tech and security firms to protect critical software using Mythos in a defensive capacity, while Anthropic funds usage credits and open-source security projects to bolster industry-wide protections.
The piece frames the NSA’s stance as part of a broader trend where governments seek the most capable AI tools despite governance and strategic risks, raising questions about trust, accountability, and how such tools are deployed in national security. 21 April 2026.