A federal court upheld the Pentagon's blacklisting of Claude

A D.C. Circuit appeals panel ruled 2-1 to uphold the Defense Department's supply-chain-risk designation against Anthropic, rejecting the company's argument that the blacklist was arbitrary and unconstitutional. Judge Gregory Katsas wrote for the majority that continued integration of Claude into Pentagon systems presented a "statutorily covered national-security risk."

A contract that never got signed

Anthropic signed a $200 million Pentagon contract in July 2025. Negotiations over deploying Claude on the DOD's GenAI.mil platform collapsed two months later, when the department asked for unrestricted access to the models for any lawful purpose. Anthropic would not grant that without restriction, specifically citing autonomous weapons and domestic mass surveillance as uses it would not clear a model for in advance.

What the ruling actually decided

The court did not rule on whether Claude is unsafe. It ruled that a vendor's refusal to grant a government customer unrestricted access to its own model is, by itself, enough to justify a national-security supply-chain designation. That is a narrower and stranger finding than a headline about "AI blacklisted by Pentagon" suggests: the risk the court found was in the contract terms, not in anything the model did.

Anthropic's next move is unclear

An Anthropic spokesperson said the company is "considering all options, including further review," which covers everything from an en banc appeal to accepting the designation and walking away from federal contracts entirely. Neither the company nor the court's opinion says which.

Why a build studio cares

Every client we build for who wants to sell into government eventually asks what access a government buyer can demand as a condition of the sale, and this ruling gives a real answer: a refusal to grant unrestricted access, even one grounded in a specific, named misuse case, can be treated as the risk itself rather than a reasonable limit. That changes the calculus for anyone negotiating a government contract for an AI product, not just Anthropic.

Next step: read CNBC's coverage of the ruling.