The U.S. Department of Defense has initiated the first steps toward designating Anthropic as a "supply chain risk entity" amid an intensifying standoff over the military use of artificial intelligence. Such designations have previously targeted Chinese companies including Huawei.
According to Axios on the 25th (local time), the Pentagon notified Boeing and Lockheed Martin to submit assessments of their dependency on Anthropic's AI model Claude. Axios reported this marks "the first step toward designating a supply chain risk entity," adding that "it is unprecedented for the Defense Department to use such measures to penalize a U.S.-led technology company on which it currently depends."
Claude is currently the only AI model operating within U.S. military classified systems. It was used in the operation to capture Venezuelan President Nicolás Maduro and may be deployed in Iranian military operations. The military is demanding that Claude's safety guardrails be removed to enable broader military applications. However, Anthropic maintains that model modifications are not possible, citing concerns over potential use in mass civilian surveillance and AI weapons development.
At a meeting on the 24th, Defense Secretary Pete Hegseth told Anthropic CEO Dario Amodei that if the company does not agree to Pentagon conditions by 5:01 p.m. on the 27th, the department would declare Anthropic a supply chain risk entity and forcibly modify Anthropic's model to meet national defense needs.
Axios noted that while Anthropic could challenge such measures through litigation, invoking the Defense Production Act would prevent the company from blocking the Pentagon's use of Claude. The outlet also raised concerns that an actual supply chain risk designation could deter many government contractors from using Claude.
Meanwhile, Elon Musk's xAI recently signed a contract with the Pentagon under terms permitting "all lawful uses"—conditions Anthropic rejected. Google and OpenAI are also reportedly in related negotiations. One source said, "Claude is currently the most capable model for military applications, but Google's Gemini is emerging as a strong alternative." However, Google and OpenAI would also likely need to agree to all lawful uses to secure contracts.