A federal judge has ruled that the Trump administration hasn’t presented enough evidence to justify labeling Anthropic a supply-chain risk, casting serious doubt on the government’s ban on the AI company’s technology.

The ruling is the latest setback in the administration’s attempt to restrict Anthropic under supply-chain security authorities. It’s also a test case for how much deference courts will give federal agencies when they try to regulate AI companies on national security grounds.

The legal standard

Supply-chain risk designations typically fall under executive orders and statutory frameworks that give agencies broad authority to restrict foreign technology or companies with foreign ties. But that authority isn’t unlimited. Agencies still need to show their work.

The judge’s ruling suggests the government hasn’t done that here. Without seeing the full opinion, the key question is what standard the court applied. If it’s arbitrary-and-capricious review under the Administrative Procedure Act, the government has a pretty low bar to clear. It just needs to show it considered the relevant factors and didn’t act irrationally.

If the government can’t even meet that standard, it means the evidence is exceptionally weak or the reasoning is fundamentally flawed.

What this means for Anthropic

The immediate effect is that the government’s ban on Anthropic’s technology remains on shaky legal ground. Depending on the procedural posture, the ban might be stayed or enjoined while the case continues.

For Anthropic, this is a partial win. But it’s not over. The ruling says the administration “still” lacks evidence, which implies the government can try again with better evidence or a stronger legal theory. Federal agencies can usually cure defects in their reasoning by going back and supplementing the administrative record.

The word “still” is doing a lot of work in that headline. It suggests this isn’t the first time the court has found the government’s case wanting. That’s significant. If the agency already had one chance to bolster its evidence and failed, judges start getting less patient.

The broader context

This case sits at the intersection of administrative law, national security, and AI regulation. All three areas tend to give the government significant discretion. National security cases especially tend to be deferential to executive branch judgments.

So when a court says the government hasn’t shown enough, even in a national security-adjacent case, that’s notable.

It also raises questions about how the government is approaching AI regulation more broadly. Supply-chain security authorities were designed for hardware and traditional software with clear foreign government ties. Applying them to AI companies requires showing similar risks exist. That’s a harder case to make, especially for a U.S.-based company like Anthropic.

What happens next

The government has several options. It can appeal, supplement the record with new evidence, or try a different legal theory entirely. Or it could drop the designation, though that seems unlikely given the political investment.

Anthropic will likely continue defending the case while trying to operate under whatever restrictions are currently in place. The company’s ability to work with government agencies, secure cloud providers, or deploy in certain sectors could all be affected by the supply-chain risk label.

The case also matters for other AI companies. If the government can’t make this designation stick against Anthropic, it suggests similar actions against other domestic AI companies would face steep legal challenges. That doesn’t mean the government won’t try, but it does mean companies have viable defenses.

The evidence question

The most important unanswered question is what evidence the government actually presented and why the court found it insufficient. Was it purely speculative? Did it rely on classified information the court couldn’t verify? Did it conflate general AI risks with supply-chain risks specific to Anthropic?

Without seeing the court’s reasoning, it’s hard to say. But the fact that a federal judge found the evidence lacking in a case touching on national security suggests the government’s case has real problems.

For anyone watching how AI regulation develops, this case is worth following. It’s one of the first major legal challenges to government restrictions on AI companies, and it’s testing how much authority agencies actually have in this space. The answer appears to be less than they thought.