According to Axios, Anthropic’s recently disclosed model, Mythos Preview, is allegedly being used by the National Security Agency. The announcement was made weeks after the Department of Defense, the parent organization of the NSA, designated Anthropic as a “supply-chain risk” due to the company’s refusal to grant Pentagon authorities unrestricted access to the full capabilities of its model.
Earlier this month, Anthropic unveiled Mythos, a frontier model intended for cybersecurity duties. However, they asserted that the model was too capable of offensive cyberattacks to be made public. Consequently, the AI company restricted access to Mythos to about forty entities, of which only twelve have been made public.
The NSA is reportedly employing Mythos mostly to scan environments for exploitable vulnerabilities, and it looks to be one of the unidentified recipients. The AI Security Institute in the United Kingdom has also acknowledged having access to Mythos.
The U.S. military is using Anthropic’s equipment more frequently while simultaneously claiming in court that it poses a threat to national security. The Pentagon’s disagreement began when Anthropic refused to provide Claude for the development of autonomous weapons and widespread domestic monitoring.
The NSA’s access to Mythos coincides with what seems to be a warming of Anthropic’s ties to the Trump administration. Secretary of the Treasury Scott Bessent and White House head of staff Susie Wiles met with Anthropic CEO Dario Amodei last Friday.
The meeting was apparently deemed fruitful by the White House.