Anthropic Listed as National Security Risk for Upholding Claude Safety Red Lines, Files for Rehearing After 2-1 Loss

Beating AI News Flash: Anthropic and the Pentagon are still locked in their legal battle. On October 13, Anthropic asked the full bench of the Washington federal appeals court to rehear the blacklist case. On September 25, a three-judge panel ruled 2-1 in favor of the Pentagon, keeping Anthropic listed as a national security supply chain risk company. Anthropic says the ban has already cost it billions of dollars in business.

The two sides were originally partners. In July 2025, Anthropic had just secured a $200 million contract with the Pentagon, but talks later fell apart over restrictions on Claude's use. Anthropic insisted on two bottom lines: AI must not be used for mass surveillance of Americans, and AI must not be allowed to autonomously decide on lethal attacks. The military, meanwhile, demanded the ability to use Claude for all lawful military purposes. In March of this year, the Pentagon blacklisted Anthropic, barring continued use of its technology in defense projects.

The dispute centers on why setting safety restrictions on AI would instead constitute a national security risk. The Pentagon argues that Claude has indeed refused to carry out government tasks in the past due to built-in restrictions. If it suddenly cannot respond to commands during wartime, it could affect military operations. Two judges accepted this reasoning. Anthropic counters that the relevant law was originally intended to guard against malicious supply chain sabotage, but is now being stretched to cover normal model safety restrictions set by companies.