Court lets Pentagon label Anthropic a supply-chain risk

A federal appeals court in Washington let the Pentagon keep Anthropic and Claude out of its supply chain, accepting the government’s argument that Anthropic’s refusal to allow some military uses of Claude can count as a national-security procurement risk.

The dispute turns on Anthropic’s limits against using Claude for fully autonomous lethal warfare and mass domestic surveillance. The Department of War wanted an “all lawful uses” term; Anthropic refused, and the department moved to exclude the company under federal supply-chain law.

The D.C. Circuit denied Anthropic’s petitions for review. The majority said the department had enough support for treating Claude’s built-in restrictions as a covered risk, even without finding malicious intent by Anthropic. It also rejected Anthropic’s free-speech and due-process claims.

The decision does not end every related fight. A separate federal case has already rejected a parallel government designation, and Anthropic told ABC News it is considering further review. But for now the ruling gives the Pentagon room to keep Claude out of defense systems while other labs compete for military AI contracts.

For AI companies, the useful signal is broader than one procurement fight: model safety restrictions are no longer only a product-policy question. In government contracts, they can become part of national-security risk analysis.