Axios reported Sunday that the United States National Security Agency is using Anthropic’s Mythos Preview AI tool, even though the Pentagon has formally flagged the company with a supply-chain risk designation.

Axios said, citing sources, that the Mythos Preview model was being used more broadly within the agency.

Last week, U.S. President Donald Trump’s administration and Anthropic’s CEO discussed potential collaboration for the first time since a dispute earlier this year between the Pentagon and the AI company over how its models should be used.

The discussions came amid rising concerns that Anthropic’s latest AI model, Mythos, could significantly enhance cyberattack capabilities.

Anthropic has previously described the model as its “most capable yet for coding and agentic tasks,” referring to its ability to operate with a degree of autonomy.

Anthropic, the NSA, and the U.S. Department of Defense did not immediately respond to requests for comment outside regular business hours.