A federal judge has ruled that the Department of Defense illegally retaliated against Anthropic for refusing to let the military use its AI models for mass surveillance of U.S. citizens.

The decision, announced Monday, found that the Pentagon’s designation of Anthropic as a “supply chain risk” violated the company’s First Amendment rights. The label was intended to punish Anthropic for telling the military it wouldn’t allow its technology to be weaponized against Americans.

According to the ruling, that constituted unlawful retaliation. The judge’s exact language: the designation “constituted unlawful retaliation in violation of the First Amendment.”

What Anthropic actually did

The case stems from Anthropic’s decision to set boundaries on how its AI models could be used by government contractors. When the DOD approached the company about incorporating its technology into surveillance systems targeting U.S. persons, Anthropic said no.

That’s protected speech. Companies can tell the government they won’t participate in programs they find objectionable. The First Amendment covers corporate speech, and it definitely covers a company saying “we’re not helping you do that.”

The Pentagon apparently didn’t see it that way. Instead of accepting the refusal, the DOD slapped Anthropic with a supply chain risk designation, a label typically reserved for companies that pose actual security threats. It’s the kind of designation that can tank government contracts and scare off private sector clients who work with federal agencies.

The court saw this for what it was: punishment for speech the government didn’t like.

Why this matters beyond Anthropic

This ruling establishes important precedent for AI companies navigating government contracts. It says clearly that refusing to participate in surveillance programs is protected speech, and the government can’t use procurement designations as a cudgel to punish companies for that refusal.

The Electronic Frontier Foundation joined a coalition of organizations in filing multiple amicus briefs supporting Anthropic’s position. EFF’s involvement signals this case has implications far beyond one company’s relationship with the Pentagon.

The broader principle at stake is whether the government can use administrative labels to chill corporate speech. If the DOD could get away with branding Anthropic a security risk for declining a contract, that’s a playbook for silencing any company that pushes back on government overreach.

The judge shut that down.

The surveillance context

The timing here is notable. The DOD’s move against Anthropic came as the Pentagon was rapidly expanding its use of AI for intelligence gathering and analysis. Multiple agencies have been pushing to integrate large language models into surveillance infrastructure, often with minimal public disclosure about capabilities or safeguards.

Anthropic’s refusal to participate in mass surveillance of Americans put the company at odds with lucrative defense contracts. But it also drew a line that apparently made the Pentagon uncomfortable enough to retaliate.

The First Amendment doesn’t just protect speech you agree with. It protects companies that tell the government no, even when that no costs money and angers powerful agencies.

What happens next

The ruling doesn’t end the case. The DOD could appeal, though it’s not clear they will. The supply chain risk designation is presumably now invalid, which means Anthropic should be able to pursue government contracts without that cloud hanging over them.

More significantly, this decision gives other AI companies a legal foothold if they face similar retaliation. The precedent is clear: the government can’t weaponize administrative designations to punish protected speech.

Whether other companies will actually follow Anthropic’s lead is a different question. Saying no to the Pentagon takes guts, especially when competitors are saying yes. But at least now there’s a court ruling that says companies have the right to make that choice without illegal retaliation.

The DOD hasn’t commented publicly on the ruling. Anthropic released a brief statement calling the decision “an important vindication of fundamental rights,” which is understated but accurate.

The bigger picture

This case sits at the intersection of three major issues: government surveillance, AI deployment, and corporate free speech. All three are contentious, and all three are only getting more complicated as AI capabilities expand.

The fact that a federal judge had to step in and tell the Pentagon it can’t punish companies for refusing surveillance work is itself revealing. It suggests the DOD believed it had the authority to do exactly that, or at least thought it could get away with it.

The court disagreed. That’s good news for anyone who thinks companies should be able to decline government contracts without facing retaliatory administrative action.

It’s also good news for the First Amendment, which apparently still applies even when the speech in question is a corporation telling the military no.