The U.S. Department of Defense has initiated a six-month transition plan to discontinue the use of Anthropic's Claude models, citing them as a "supply chain risk." This move is part of a broader strategy to diversify its AI supplier base. The Pentagon conducted a "super user" test with 25 AI specialists to evaluate alternatives, focusing on logical reasoning, data security, and execution speed in classified missions. In response, the Pentagon has established partnerships with multiple leading AI companies to deploy a variety of AI tools on classified networks, aiming to build a resilient AI application ecosystem. This strategic shift underscores the importance of "supply chain security" and "vendor background" in government-level AI procurement, moving beyond mere model performance considerations.