Federal Judge: Pentagon's Anthropic Ban Looks Like Punishment

Lin said she was concerned that the government was punishing Anthropic for going public with its dispute over Claude's use in autonomous weapons and domestic surveillance — a constitutional issue, not just a contractual one.

A ruling is expected within days on whether to temporarily pause the Pentagon's ban while the courts decide on the merits of the case.

To recap the timeline:

Late February 2026: Anthropic CEO Dario Amodei publicly announced Claude would not be used for fully autonomous weapons or to surveil American citizens — two carve-outs the Pentagon had been pushing to remove from the contract.

Shortly after: President Trump ordered all U.S. government agencies to stop using Anthropic products.

Early March 2026: The Pentagon formally designated Anthropic a "supply chain risk" under 10 USC 3252 — a designation historically used against foreign adversaries, never before applied to a U.S. company over a policy dispute.

Anthropic's response: Two federal lawsuits. One filed in the Northern District of California (where this week's hearing took place), one in the D.C. federal appeals court. Both allege First Amendment violations and statutory overreach.

Anthropic's core claims:

First Amendment retaliation. Amodei made a public statement about Claude's use limits. Days later, the government banned all federal use of Claude. Anthropic argues this sequence is textbook retaliation for protected speech.

Statutory overreach. The supply chain risk designation statute, Anthropic argues, was designed for foreign adversaries posing sabotage risks — not for American companies that disagree with government contract terms. Using it this way exceeds the law's scope.

Collateral damage. The designation doesn't just cut off DoD contracts. By threatening Pentagon contractors with consequences for doing business with Anthropic, it effectively extends the blacklist to the broader economy in ways the statute doesn't authorize.

Judge Lin acknowledged the government's right to choose its AI vendors. But she drew a distinction between "we don't want to use your product" and "we're going to designate you a national security threat" — the latter being an action with far broader legal and reputational consequences.

If granted, it would pause the supply chain risk designation while the full case proceeds. That matters because:

Anthropic's customers — particularly contractors with DoD exposure — are making decisions right now about whether to continue their Claude relationships. Every week the designation stands, some of those customers move on.

The designation's reputational impact compounds over time. A pause would limit damage while the underlying question gets resolved.

Anthropic has argued irreparable harm — the kind of ongoing injury courts consider when deciding whether to freeze government action pending trial.

The question in front of Judge Lin is narrow: should the designation be paused temporarily? But the broader question is bigger.

Can the executive branch use supply chain risk law to punish U.S. companies for AI safety policies the administration dislikes? If yes, every AI company with government revenue now has to weigh whether their safety commitments are negotiable under political pressure.

If Anthropic wins on the First Amendment claim, it would establish that AI safety policies — at least those publicly stated — are protected speech, and that government retaliation for expressing them is unconstitutional.

That's a significant precedent for an industry that's increasingly having to navigate between commercial, government, and safety imperatives simultaneously.

The hearing changed nothing about Claude access. Individual users, enterprise API customers, and non-DoD organizations are unaffected. The ruling — expected shortly — will determine whether the designation is paused or allowed to proceed.

We'll update when Judge Lin's decision comes down.