Anthropic PBC v. U.S. Department of War (Anthropic N.D. 2026)
Case Overview
Anthropic sued after refusing to remove two usage restrictions on Claude: one prohibiting lethal autonomous drone operations with no human oversight, and one blocking mass surveillance of Americans. Hegseth designated Anthropic a 'Supply-Chain Risk to National Security' — the first American company ever to receive this designation — and cancelled all government contracts. Anthropic filed five counts: APA violations, First Amendment retaliation, ultra vires executive action, Fifth Amendment due process, and additional APA sanction violations. Judge Rita Lin granted a preliminary injunction March 26, 2026, finding defendants 'failed to prove' the supply-chain risk designation and a 'high likelihood of success' on the First Amendment retaliation claim.
The Facts
Defense Secretary Hegseth directed Anthropic to remove two usage restrictions from Claude: one prohibiting use for lethal autonomous weapons without human oversight, and one prohibiting mass surveillance of Americans. Anthropic refused on safety grounds. Hegseth then formally designated Anthropic a Supply-Chain Risk to National Security, the first such designation of an American AI company, effectively barring it from government contracts.
The Application
Hegseth's demand that Anthropic remove safety restrictions on Claude constitutes government compulsion of private speech—an attempt to coerce removal of the company's own product policy. The government's response—designating Anthropic a supply-chain risk and canceling contracts—uses federal procurement authority as a retaliatory enforcement mechanism, violating the rule that government cannot leverage contracting power to punish constitutionally protected product decisions. Judge Lin found that Anthropic has a high likelihood of success because the timing and sequence (demand → refusal → designation) demonstrate retaliatory intent, and the designation appears motivated by viewpoint disapproval rather than independent national security analysis.
The Conclusion
PENDING. If the court enjoins the designation, Anthropic's safety restrictions on Claude stand and the government cannot use national security labels to coerce removal of AI guardrails. If upheld, the executive may use procurement designations to pressure AI developers over product safety policies.
Flag an issue
This tracker is maintained by BrynoDC and is free because readers fund it. Support