NSA Deploys Anthropic's Claude Mythos on Classified Networks Amid Pentagon Legal Battle

The National Security Agency (NSA) is reportedly running Anthropic's powerful Claude Mythos Preview AI model on its classified networks. This operational deployment occurs against a backdrop of significant institutional friction, as the Department of Defense (Pentagon) is simultaneously engaged in a legal battle against the AI company. The dual-track approach—embracing the technology for sensitive intelligence work while another major defense arm pursues litigation—highlights the complex and often contradictory pressures within the U.S. national security establishment regarding cutting-edge artificial intelligence.

The specific model in use, Claude Mythos Preview, represents Anthropic's most advanced and capable system to date, designed for complex reasoning tasks. Its integration into the NSA's secure infrastructure signals a strategic bet on the model's utility for intelligence analysis and data processing at the highest classification levels. This move comes as Anthropic's CEO, Dario Amodei, has been meeting with White House officials, underscoring the company's deepening engagement with the highest levels of the U.S. government even as it faces legal challenges from the Pentagon.

The situation creates a stark contrast: one arm of the U.S. security state is actively leveraging Anthropic's technology for potential operational advantage, while another is contesting the company's actions in court. This divergence points to unresolved tensions over AI governance, procurement, and risk assessment within the defense and intelligence communities. The deployment on classified networks also raises immediate questions about the security protocols, data handling, and oversight mechanisms governing the use of such powerful commercial AI systems in the nation's most sensitive environments.