The Pentagon is currently assessing artificial intelligence models from various providers as it seeks to replace Anthropic’s Claude across its operations.
In July 2025, Anthropic secured a substantial two-year contract valued at $200 million. The goal was to seamlessly integrate Claude into classified defense networks. However, complications began to emerge in January 2026 when Claude was allegedly deployed in a military operation aimed at capturing Venezuelan President Nicolás Maduro. This usage highlighted tensions, as it contradicted Anthropic's requirements concerning lawful applications, leading to friction with the Department of Defense.
Anthropic had established firm boundaries regarding acceptable AI applications, particularly rejecting fully autonomous weapon systems and widespread surveillance tools. By late February 2026, Defense Secretary Pete Hegseth issued a firm ultimatum to Anthropic. When the company stood by its principles, Hegseth classified the organization as a supply chain risk. This designation kicked off a systematic transition away from Claude, aiming for a complete changeover within a three to twelve-month timeframe.
In response to this situation, the Pentagon has hastened the implementation of alternative AI solutions from OpenAI and Elon Musk’s xAI, the creator of Grok. OpenAI revisited its agreement with the Department of Defense, ensuring unimpeded lawful use of its AI technologies, in stark contrast to the demands Anthropic was unwilling to meet.
Complicating matters, it was revealed that the NSA had been testing Anthropic's Claude Mythos preview on classified networks as late as April 2026, despite the Pentagon's ongoing efforts to phase out Claude altogether.