Policy

Appeals Court Upholds Pentagon Ban on Anthropic's Claude

A federal appeals court upheld the Pentagon's supply-chain risk label on Anthropic, allowing the military to block Claude models after the startup refused to support autonomous warfare.

WIRED AI1 day agoPolicy
Image: WIRED AI

A federal appeals court in Washington, D.C., has ruled 2-1 against Anthropic, upholding a Pentagon designation that labels the artificial intelligence startup a national security supply-chain risk. The decision allows the Department of Defense to continue its efforts to phase out Anthropic's Claude models from military and federal systems. The ruling counteracts a separate decision by a federal judge in San Francisco, who threw out a different supply-chain label against the company in March.

The legal battle stems from Anthropic's policy prohibiting the deployment of its current AI models for autonomous weaponry or domestic surveillance operations. Defense Secretary Pete Hegseth deemed this restriction a national security risk, leading the Pentagon to sanction Anthropic under two separate supply-chain laws. In its decision, the appeals court majority rejected Anthropic's claims of free speech and due process violations, stating the government excluded the firm for refusing to agree to an essential contract term.

For AI practitioners and enterprise customers, the ruling highlights the growing tension between ethical safety guardrails and lucrative government contracts. While Anthropic has reported lost revenue from clients wary of doing business with a sanctioned entity, the company is still preparing for a potential initial public offering later this year. Meanwhile, the Pentagon is actively looking to replace Claude with alternative models, including SpaceX's Grok, Google's Gemini, and OpenAI's GPT models.

This split legal landscape means Anthropic faces years of ongoing litigation, with spokesperson Danielle Cohen stating the company is considering an appeal to the Supreme Court. For developers and tech firms, the case establishes a precedent that the U.S. military can legally blacklist AI vendors that refuse to modify their safety protocols for defense applications, forcing other major AI labs to choose between internal ethical guidelines and massive federal defense integration.

This is our own summary of reporting by WIRED AI

More in Policy