The Trump administration has had months to build its case against Anthropic. According to the federal judge overseeing that case, it's gotten weaker, not stronger, in the meantime. At a hearing in San Francisco on July 30, 2026, U.S. District Judge Rita Lin said the government still hasn't presented enough evidence to justify designating Anthropic a supply-chain risk, the label underpinning a broader federal effort to cut the AI company off from government contracts entirely.

This is a different fight from the 90-minute shutdown order that pulled Anthropic's Fable 5 model offline back in June. That earlier action targeted a specific model under export control law after a jailbreak discovery. This case is bigger and more structural: it's about whether the federal government can formally blacklist Anthropic as a company, based on a national security label, and whether that label was ever actually backed by evidence rather than frustration over a stalled contract negotiation.

How This Started: A Contract Dispute Over Autonomous Weapons

The root of the conflict traces back to negotiations between Anthropic and the Department of Defense that broke down earlier this year. Anthropic wanted contractual limits ensuring its AI wouldn't be used for mass surveillance of Americans, or for targeting and firing decisions involving lethal autonomous weapons, arguing the technology simply wasn't ready for those applications. The Pentagon pushed back, insisting a private company shouldn't get to dictate how the military uses the tools it buys, and said it intended to use Anthropic's technology only in ways it considered lawful.

When those talks collapsed, the administration didn't just walk away from the deal. It designated Anthropic a supply-chain risk and directed federal agencies to immediately cease all use of the company's technology, a step Anthropic's lawsuit called an unprecedented and unlawful attempt to punish the company for First Amendment-protected speech.

The Order That Went Further Than the Pentagon

Judge Lin's earlier ruling in this case, issued after a first hearing, already found real problems with how far the administration's order reached. Defense Secretary Pete Hegseth had posted publicly that military contractors must cut off all commercial activity with Anthropic, not just military-related work. Lin formally rejected that directive, finding it appeared to illegally require companies to stop using Claude even for entirely unrelated, non-military business. She also noted that after the administration's initial action, federal agencies well outside the Pentagon quickly terminated their own use of Claude on their own, a ripple effect that put Anthropic's broader public sector business at risk over a dispute that originated in a single stalled defense contract. One amicus brief filed in the case described the government's measures bluntly as attempted corporate murder.

That earlier ruling was temporary, giving the government a week to appeal before it took full effect. The administration did appeal. This week's hearing was about whether Lin's temporary block should become permanent.

What the Government Actually Has, and Doesn't

At Thursday's hearing, the case for the government didn't improve. According to Startup Fortune's reporting, Lin found no evidence supporting claims about kill-switch capabilities or China-linked security threats, the kind of concrete technical findings that would typically justify a formal supply-chain risk designation. Notably, a Justice Department attorney, Eric Hamilton, conceded during the hearing that even a valid supply-chain risk designation would only justify blocking contractors from using Anthropic's technology for military-related work specifically, a considerably narrower scope than the sweeping, government-wide cutoff the administration originally ordered.

That concession matters. It suggests even the government's own legal team is backing away from the broadest version of the ban it initially pursued, while the underlying evidentiary basis for any version of the designation remains, in the judge's own assessment, unproven.

The Argument the Judge Called "Really Troubling"

Perhaps the most consequential moment from Thursday's hearing involved a specific argument the government made in its own defense: that Anthropic's public criticism of the Department of Defense itself justified the ban. Lin called that reasoning really troubling, warning explicitly that accepting it could set a precedent allowing the government to retaliate against any federal contractor that publicly disagrees with the administration's policies. That's a significant legal line to flag, since it reframes this case from a narrow dispute about one company's AI safety guardrails into a broader test of whether federal contractors retain First Amendment protection when they push back publicly against how the government wants to use their technology.

What's Actually at Stake for Anthropic Right Now

Lin hasn't yet issued a final ruling making her earlier block permanent, but her comments this week make her skepticism of the government's position hard to miss. For Anthropic, the practical stakes go well beyond legal principle. The company's public sector business, which had been growing as government agencies adopted Claude across multiple departments, was thrown into uncertainty the moment the original designation was announced, and Anthropic has said some government contractors remain worried about running afoul of the president's order even now, during the period the block has been in effect.

This case is also unfolding against the backdrop of Anthropic's broader push toward a public offering, at a moment when the company's revenue and enterprise relationships are under closer scrutiny than ever. A federal ban, even one a court eventually strikes down permanently, creates exactly the kind of prolonged uncertainty that complicates due diligence for investors evaluating a company ahead of a major IPO. Whatever Lin ultimately decides, this week's hearing made one thing clear: months into the dispute, the government still hasn't shown the court the evidence it needs to make its case stick.