For six weeks, President Donald Trump’s administration has waged a bare-knuckled campaign to force Anthropic to its knees. In a move with no known precedent, the Pentagon on March 5 branded the frontier artificial intelligence company a national security supply chain risk, a designation previously reserved for foreign adversaries and Chinese tech firms like Huawei. The stated offense was that Anthropic refused to let the U.S. military use its AI systems for fully autonomous weapons or the mass surveillance of American citizens. Trump described the decision with his trademark directness, saying he fired Anthropic “like dogs.”
As it turns out, dogs bite back. Anthropic is fighting the government’s designation in two separate federal courts on two separate legal tracks. In late March, a judge in San Francisco called the Trump administration’s retaliatory response “Orwellian” and granted Anthropic’s request for a preliminary injunction. In a separate legal decision in early April, an appeals court in Washington conceded that Anthropic faced real harm from the Pentagon’s designation but said the financial stakes were not yet clear enough to warrant blocking the move outright. A final resolution of Anthropic’s status could take weeks if not months.
While the lawyers argue, the National Security Agency has already reached its own verdict. It has reportedly begun running Claude Mythos Preview, a model Anthropic considers too dangerous for public release, on its own systems. The message is plain: Whatever Anthropic is, it is too important to sideline.