The Pentagon has entered a dispute with San Francisco-based artificial intelligence company Anthropic over the use of its flagship AI model, Claude. The disagreement centers on Anthropic’s refusal to allow the Department of Defense unfettered access to Claude without security restrictions for all lawful purposes. This conflict has led the Pentagon to consider cutting business ties with the company amid ongoing negotiations that defense officials describe as difficult and protracted.
Defense Secretary Pete Hegseth moved closer to designating Anthropic as a supply chain risk, which would mark a significant step in unraveling the collaboration between the U.S. military and the AI firm. The designation would reflect concerns about the security and operational reliability of Anthropic’s AI model in sensitive defense applications. Pentagon officials have publicly noted frustration with Anthropic’s position after months of efforts to reach an agreement.
Anthropic’s Claude model is currently integrated into classified military networks, forming a critical component in some defense operations. The Pentagon’s review comes as the department evaluated the company’s role in a recent high-profile raid related to Venezuelan President Nicolás Maduro. Despite the use of Anthropic’s AI during that operation, the ongoing contract faces thorough scrutiny as the military reassesses its security posture concerning third-party technology providers.
Chief Pentagon spokesman Sean Parnell confirmed that the department’s relationship with Anthropic remains under review but did not provide details on a timeline or potential outcomes. The dispute highlights the broader challenge the U.S. military faces in balancing cutting-edge AI innovation with stringent security requirements. The Pentagon’s reliance on advanced AI models like Claude reflects growing efforts to enhance operational capabilities, yet the standoff with Anthropic demonstrates the complexities inherent in managing partnerships with private technology firms.
Anthropic declined to remove the security limitations the Pentagon requested, citing concerns over the unrestricted use of its AI model. The company’s refusal has prompted defense officials to consider the potential risks associated with continued collaboration. As of mid-February 2026, no public resolution had emerged, and discussions remained tense while the Pentagon weighs next steps regarding its AI strategy and vendor relationships.
