The NSA is reportedly using Anthropic's new Mythos model

April 20, 2026

An article summarized by Axios:

National Security Agency is reportedly using Anthropic’s advanced AI model, Mythos Preview, even as the Department of Defense/ War has labeled the company a potential “supply chain risk.” This creates a contradiction, as the Pentagon has been trying to restrict Anthropic’s role while parts of the U.S. government continue adopting its technology for cybersecurity and intelligence purposes.

The model is believed to be used for identifying vulnerabilities in systems, a powerful capability that Anthropic has limited to a small group of organizations due to its potential misuse. While details on the NSA’s exact use remain unclear, sources suggest broader adoption within defense-related agencies. At the same time, Anthropic executives have been meeting with top government officials to discuss how the technology can be safely integrated across agencies.

The tension stems from a deeper dispute over how Anthropic’s AI should be used. The Pentagon has pushed for fewer restrictions, including broader military applications, while Anthropic has resisted uses tied to mass surveillance or autonomous weapons. Despite ongoing legal and policy battles, some officials want to move past the conflict to take advantage of cutting-edge AI tools, highlighting the growing importance of advanced AI in national security.