A federal intelligence agency continues to employ an artificial intelligence system from a company the Pentagon has flagged as a potential security concern, according to a weekend report from Axios.
Sources tell the news outlet that the National Security Agency has been utilizing Anthropic’s Mythos Preview AI technology, even after the Defense Department issued a formal supply-chain risk classification against the firm. The intelligence agency falls under Defense Department oversight.
According to the reporting, usage of the Mythos Preview system has expanded throughout the department. Reuters noted they were unable to independently confirm these details.
Representatives from Anthropic, the NSA, and the Defense Department have not yet provided responses to media inquiries made during weekend hours.
Just last week, officials from President Donald Trump’s administration held their first discussions with Anthropic’s chief executive about potential collaboration since tensions arose earlier this year between Pentagon officials and the AI company regarding proper usage protocols for the firm’s technology.
These conversations occurred as cybersecurity specialists express mounting concerns that the artificial intelligence startup’s newest Mythos model could amplify the threat of digital attacks.
Anthropic has described the system as their “most capable yet for coding and agentic tasks,” highlighting the technology’s capacity to operate independently.
Security experts warn that the model’s sophisticated programming abilities could give it extraordinary power to detect weaknesses in cybersecurity systems and develop methods to exploit those vulnerabilities.
