Back in March, a federal judge blocked the Pentagon from labeling Anthropic a “supply chain risk” and banning federal use of its AI. The government had argued Anthropic was untrustworthy enough to disable its own models mid mission, among other claims, and the judge wasn’t buying it even then.

On Thursday, the same judge, Rita Lin, told the government its case had actually gotten worse since March, not better. She said she still hasn’t seen evidence that Anthropic can disable or alter a model it has already delivered to a customer, no kill switch, nothing that backs up what the Pentagon claimed in its original directive.

The more uncomfortable part of the hearing was about why the ban happened in the first place. Anthropic had publicly pushed back on how the Pentagon wanted to use its AI, saying no to mass surveillance work and no to letting a model make lethal targeting calls on its own. The government argued that public pushback was itself justification for the ban, that criticizing the customer was reason enough to blacklist the vendor.

Judge Lin didn’t accept that. She called the argument “really troubling” and said it runs up against the First Amendment, warning it could tell every federal contractor that disagreeing with the administration in public comes with a price tag attached.

The case isn’t over yet, and the government could still come back with more. But it’s a good sign for anyone who wants AI companies to be able to say no to a use case without getting blacklisted for it. Anthropic drew a line on surveillance and lethal weapons, and so far, that line is holding up in court. That’s worth rooting for, whichever company draws the next one.

——

Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277

——

Sources:
https://techcrunch.com/2026/07/30/judge-says-trump-admin-still-lacks-evidence-for-anthropic-supply-chain-risk-label/
https://www.cbsnews.com/news/anthropic-ruling-judge-trump-pentagon-ai/
https://www.courthousenews.com/judge-likely-to-rid-anthropic-of-pentagons-supply-chain-risk-label/

Repost this. Thanks.