Court rejects Anthropic's appeal

A federal appeals court in Washington left Anthropic’s blacklisting in place on April 8, refusing to pause the Pentagon’s decision to label the company a “supply chain risk” while the case continues. The order means defense contractors still cannot use Claude on Department of Defense work for now. (cnbc.com) The judges said Anthropic would likely suffer some harm, but they weighed that against the government’s control over how it buys “vital” artificial intelligence tools during an active military conflict. In the court’s framing, lost revenue for one company did not outweigh the Pentagon’s procurement authority. (politico.com, cnbc.com) This fight started in late February, when Defense Secretary Pete Hegseth gave Anthropic a deadline to allow broader military use of Claude. Dario Amodei, Anthropic’s chief executive, said he would not permit uses tied to mass surveillance of Americans or fully autonomous weapons. (euronews.com) Anthropic was not a fringe defense vendor before this clash. The company said it had won a $200 million Defense Department contract in July 2025 to prototype frontier artificial intelligence for national security, and it had also partnered with Palantir to place Claude inside intelligence software. (euronews.com, euronews.com) A “supply chain risk” label is procurement law’s version of a no-fly list for vendors. Once the Pentagon applied it in early March, contractors working on Defense Department projects had to certify they were not using Anthropic’s models in that work. (euronews.com, cnbc.com) That label reaches farther than one canceled contract. Lockheed Martin said it would follow the order and look for other large language model providers, a sign that a procurement ruling can push a startup out of an entire defense ecosystem, not just out of one office. (euronews.com) Anthropic did win one important ruling in California on March 27. Judge Rita Lin temporarily blocked enforcement of the broader government ban and wrote that the statute did not support branding an American company a national security threat for disagreeing with the government. (euronews.com) But United States procurement law split the case in two places at once. Even after the California injunction, Anthropic still had to ask the District of Columbia Circuit to stop the designation itself, and that is the request the appeals panel just denied. (politico.com, euronews.com) So Anthropic is now in a narrow middle ground. It can keep working with other federal agencies while the lawsuits continue, but it is still shut out of Defense Department contracts and its tools are still off-limits for Pentagon contractors. (cnbc.com) The case is turning into a test of how much leverage the government has over artificial intelligence companies that want public-sector business but also want to draw lines around military use. Anthropic told the court the designation was unconstitutional retaliation and unlawful procedure; the court said those merits questions can move quickly, but not quickly enough to lift the label now. (cnbc.com, politico.com)