Federal Judge Blocks Pentagon From Labeling Anthropic a Security Threat

A U.S. District Court judge in San Francisco has issued a preliminary injunction blocking the Pentagon and the Trump administration from enforcing a national security designation against Anthropic and restricting its Claude AI for federal agencies. On March 26, U.S. District Judge Rita F. Lin ruled the government’s actions likely violated the First Amendment, failed due process, and exceeded authority under the Administrative Procedure Act. The injunction is stayed for seven days, giving the administration time to seek an emergency appeal to the Ninth Circuit by about April 2. The dispute started when the Department of Defense pushed for unrestricted access to Anthropic’s Claude models for federal use. Anthropic kept safety guardrails in its acceptable-use policy, including limits on mass domestic surveillance of Americans and lethal autonomous weapons without meaningful human oversight. The DoD demanded those restrictions be removed. After negotiations broke down, President Trump posted on Truth Social on Feb. 27 directing federal agencies to halt Anthropic technology use with a six-month phase-out. Defense Secretary Pete Hegseth then announced a supply-chain risk designation under 10 U.S.C. § 3252, a statute previously used for foreign adversaries, labeling Anthropic as a potential risk of “sabotage” and “subversion.” Judge Lin said the move appeared punitive and retaliatory, pointing to internal references to Anthropic’s “rhetoric” and “strong-arming.” The court also found the government had not previously applied the designation to an American company in similar circumstances, despite prior vetting (including Top Secret clearance processes, FedRAMP authorization, and contracts up to $200 million). Lin ordered restoration of the status quo, allowing existing federal contracts and partnerships to continue. The merits dispute is not resolved, and a separate challenge remains pending in the D.C. Circuit.