NSA Defies Pentagon Ban to Leverage Anthropic’s Powerful AI Model, Mythos

The National Security Agency (NSA) is utilizing Anthropic's latest AI model, Mythos Preview, despite a Department of Defense blacklist that deems the company a "supply chain risk." This decision underscores the government's prioritization of cybersecurity needs over internal disputes. The Pentagon had previously sought to restrict Anthropic's access, but the NSA's adoption of Mythos suggests a broader acceptance within federal agencies.

Anthropic has limited access to Mythos to around 40 organizations, citing security concerns due to its advanced capabilities. While the company has publicly named only 12 entities with access, sources indicate the NSA is among the unnamed users. The model is primarily employed to identify security vulnerabilities, aligning with critical defense objectives.

The situation gained momentum after Anthropic CEO Dario Amodei met with White House officials to discuss Mythos' government use and future plans. This meeting aimed to address how other departments can engage with the tool beyond Pentagon restrictions. Despite ongoing tensions, both sides viewed the discussion as productive, highlighting the administration's desire to leverage cutting-edge technology while navigating contractual complexities.

The conflict between the Pentagon and Anthropic stems from tense contract negotiations, where the Department demanded unrestricted access to Anthropic's models for "all lawful purposes." Anthropic, however, refused, citing ethical concerns over surveillance and autonomous weapons. The outcome of these discussions will shape the future of AI integration in U.S. national security efforts.