Judge Lin Says Pentagon's Case Against Anthropic 'Got Worse'

Federal judges rarely tip their hand in opening remarks at a summary-judgment hearing, which is why the tone of Thursday's proceeding in Anthropic's fight with the Pentagon is worth paying attention to. According to Axios, U.S. District Judge Rita Lin told the courtroom that the government's record 'in some ways, has gotten worse' since she first weighed the case, and said she did not see evidence that Anthropic could 'alter the model after it was delivered or flip some kind of kill switch.'

The dispute traces back to a disagreement over how Claude could be used. Anthropic restricted two categories of government use, mass surveillance of Americans and fully autonomous weapons, and the Defense Department, which wanted the model available for 'all lawful purposes,' responded by designating the company a national security and supply-chain risk. Anthropic sued. Lin previously called the administration's treatment 'troubling' and temporarily blocked the ban, and Thursday's hearing was the summary-judgment round where both sides asked her to end the case in their favor.

Why this matters beyond one contract: a supply-chain risk label is a heavy administrative weapon. It reaches across procurement, and once it sticks it is hard for a vendor to work around. If Lin agrees the government cannot support the designation on the record it has assembled, that sets a marker for what evidence future designations of frontier AI vendors actually need to survive judicial review. If she rules the other way, the ceiling on what agencies can do with that label goes up.

Take the specifics as reported, not settled. Lin is not expected to rule from the bench, she typically issues written decisions after hearings, and the losing side is expected to appeal, so anything you read into her opening remarks is a signal about where she is leaning, not the outcome. What the reporting does not give you is the concrete evidence the government submitted this round that she found unpersuasive, or how narrowly a decision for Anthropic would be written, and both of those matter for whether other AI vendors could use this ruling as cover.

The upside case, if Anthropic prevails, is a straightforward one for frontier model providers: a real precedent for pushing back on federal designations they view as retaliation for use-policy disagreements rather than a technical risk finding. That would shift the negotiating dynamic on every contract that follows.