Court won’t pause Anthropic blacklisting

A federal appeals court in Washington just left Anthropic on the Pentagon’s “supply chain risk” list, even while a judge in California had already blocked that same crackdown on March 26. That means two federal courts are now sending opposite signals about whether contractors can use Anthropic’s Claude model on Defense Department work. (cnbc.com) The Washington panel did not say Anthropic would win or lose the full case. It only said Anthropic had not met the high bar for an emergency pause, and it said the government’s wartime contracting interests outweighed the company’s immediate business losses. (politico.com) The practical effect is narrow but painful. Defense contractors can still use other artificial intelligence tools, but they cannot use Claude on Pentagon contracts while the designation stays in force. (cnbc.com) This fight started after Anthropic tried to keep Claude out of two military uses: domestic mass surveillance and fully autonomous weapons. The Pentagon argued it needed authority to use artificial intelligence for “all lawful purposes,” and the relationship broke down from there. (cbsnews.com) On March 9, Anthropic filed two cases at once: one in the Northern District of California and one in the United States Court of Appeals for the District of Columbia Circuit. That split happened because different parts of the government action had to be challenged in different courts. (courtlistener.com, courtlistener.com) The California judge, Rita Lin, called the government’s moves likely unlawful and said they looked designed to punish Anthropic for protected speech. Her March 26 injunction blocked the administration from enforcing the designation and from carrying out a broader order to stop federal use of Anthropic technology. (cbsnews.com, justia.com) The Washington court looked at the same standoff through a different lens. Its judges wrote that Anthropic would likely suffer some irreparable harm, but they treated that harm as mainly financial and said the courts should be cautious about stepping into Defense Department procurement during an active military conflict. (cnbc.com, politico.com) One detail lawyers keep circling is how unusual this label is. Politico reported that the “supply chain risk” designation had never before been applied to an American company, which is one reason this case is being watched far beyond Anthropic. (politico.com) The appeals court also fast-tracked the case, which is the court’s way of saying this fight cannot sit on a shelf for six months. So Anthropic lost the quick round, but it won a faster path to a full ruling on whether the Pentagon can use vendor-risk powers this way at all. (politico.com, techxplore.com) For procurement teams, the lesson is blunt: an artificial intelligence vendor can now become a live legal and policy risk overnight, even if the product still works the same on Tuesday as it did on Monday. Until one court clearly wins this tug-of-war, contractors, universities, and labs have to treat vendor status the way they treat export controls or sanctions: as something that can shut down a contract before any engineer writes a line of code. (computerworld.com, politico.com)