# A federal judge says the Trump administration still hasn't proven its case for restricting Anthropic > Source: > Published: 2026-07-31 05:51:33+00:00 *U.S. District Judge Rita Lin told the government this week that its evidence for branding Anthropic a national security supply-chain risk has, if anything, gotten worse since the case began.* The hearing in San Francisco on Thursday didn't go well for the Justice Department. Judge Lin opened by telling lawyers for the government that she didn't see "additional evidence really justifying what it did." Then she made it sharper: "if anything, it seems like the record has gotten worse for the government." As TechCrunch reported, she saw no proof that Anthropic could alter its Claude models after delivery, flip a kill switch, or seed them beforehand with a hidden control mechanism. No specific reason to suspect the company would try any of those things, either. That's the factual core of the administration's supply-chain risk designation. Right now it's standing on air. The government's theory, for those just catching up, is that Anthropic poses a China-linked threat to the U.S. defense industrial base. The Pentagon designated the company a supply-chain risk and issued a directive cutting it off from federal work. Anthropic sued, arguing the designation was retaliation for the company's refusal to let the DOD deploy its AI models for mass domestic surveillance of Americans or for autonomous lethal targeting. The company held that its technology wasn't ready for those uses. The Pentagon responded that a private contractor has no standing to dictate how the military uses its tools. Judge Lin granted Anthropic a preliminary injunction back in March, calling the administration's approach "Orwellian" and finding that branding an American company a potential adversary for publicly disagreeing with the government had no grounding in the governing statute. The administration appealed that ruling to the Ninth Circuit. That's where the litigation is now paused, pending Thursday's proceedings at the district level. Both sides have asked Lin to settle the case in their favour permanently, which means the judge is now weighing whether to make her preliminary block stick or dissolve it. On the China evidence specifically, she hasn't seen it. The administration has pointed to Anthropic's investors and supply chain ties as the basis for the risk label. But the kill-switch argument - the idea that a foreign actor could flip a control hidden inside a delivered Claude model - has apparently not been backed with anything Judge Lin finds credible. Frankly, for a designation this severe, that gap is remarkable. The government is asking a federal court to let it blacklist one of the leading AI companies in the country from the entire defence industrial base, and the judge is telling it, publicly and on the record, that the evidentiary foundation isn't there. Lin also called the government's argument that Anthropic's public criticism of DOD policy justifies the ban "really troubling," warning it risks establishing a precedent where any federal AI contractor that objects to government directives faces retaliation dressed up as a security designation. That's not a narrow procedural concern. If that logic were to hold, it would fundamentally reshape how AI companies negotiate the terms under which their models get deployed by the military, because the penalty for pushing back would be existential. ## What this means for enterprise AI buyers For companies and government agencies that have built workflows on Claude, the case has been a months-long source of uncertainty. The preliminary injunction issued in March temporarily unblocked Anthropic from federal work. But the designation itself was reinstated when the appeals court declined to freeze it while the government's appeal proceeded. The D.C. Circuit separately denied Anthropic's request to halt the labelling, which means two courts at once have been pulling in different directions. A permanent ruling from Judge Lin, if it goes Anthropic's way, would be the clearest resolution yet - though the Ninth Circuit appeal would almost certainly continue. The broader precedent question is what enterprise technology buyers should be watching. The government's theory - that a company's stated ethical constraints on weapon deployments can be recast as a national-security threat - has implications well beyond Anthropic. Other AI vendors, including those with DOD contracts or ambitions toward them, would be watching to see whether publicly drawing a line on autonomous weapons or surveillance becomes a liability. Judge Lin's skepticism doesn't make that risk disappear. But it does mean the government has to clear a much higher bar to make it stick. A decision isn't expected immediately. The Ninth Circuit appeal remains pending, and the district court case is now moving toward a final merits ruling. What Thursday's hearing confirmed is this: as of late July 2026, after months of litigation and appeals, the administration has not produced the evidence that would turn a politically charged security label into a legally defensible one. The judge said so herself, in open court, on the record. **Also read:** [Henry AI raises $16.5 million to become the operating system for commercial real estate deals](https://startupfortune.com/henry-ai-raises-165-million-to-become-the-operating-system-for-commercial-real-estate-deals/) • [Peter Steinberger called out Anthropic for posting ARC-AGI-3 results that OpenAI quickly made look absurd](https://startupfortune.com/peter-steinberger-called-out-anthropic-for-posting-arc-agi-3-results-that-openai-quickly-made-look-absurd/) • [Xscape Photonics raised $81 million to replace copper wiring with light inside AI data centers](https://startupfortune.com/xscape-photonics-raised-81-million-to-replace-copper-wiring-with-light-inside-ai-data-centers/)