A federal judge is openly questioning whether the U.S. government had any real basis to blacklist Anthropic as a national-security and supply-chain risk. What began as a contract dispute is now looking more like a test of how far Washington can go in punishing an AI company that resists military demands.

At the center of the fight is Anthropic’s refusal to let its models be used for two categories of work: mass surveillance of Americans and fully autonomous weapons. The Pentagon’s position is blunter: if a tool is lawful, the military should be able to use it. That clash over ethics versus state power has now landed squarely before the courts.

Judge Rita Lin sounded deeply unconvinced by the government’s evidence. In the latest hearing, she said, “I don't see additional evidence from the government really justifying what it did. If anything, it seems like the record, in some ways, has gotten worse for the government.”¹ She also rejected one of the government’s more dramatic claims — that Anthropic might tamper with its models after delivery — saying, “I don't see evidence that Anthropic could alter the model after it was delivered or flip some kind of kill switch.”²

That skepticism matters because the administration’s rationale appears to extend beyond technical risk. The government has argued that Anthropic’s public criticism of the Defense Department helps justify the restriction, a theory Lin has already described as “really troubling” because it risks turning procurement power into retaliation against dissenting contractors.³

Anthropic says its limits reflect safety and readiness concerns, not anti-military grandstanding. The Pentagon says private companies should not dictate battlefield policy. But at this stage, the court appears focused on a simpler question: whether the government can slap a sweeping risk label on a major AI firm without proving the risk is real. Right now, that case looks thin at best.⁴