Anthropic, DOD clash in DC court over blacklisting

A D.C. appeals court heard Anthropic’s challenge to the Pentagon’s blacklist of Claude models.

Anthropic and the U.S. Department of Defense faced off Tuesday in Washington, D.C., over the Pentagon’s decision to label the AI company a supply chain risk. The case centers on whether the department can keep defense contractors from using Anthropic’s Claude models while the lawsuit continues.

What changed

The appeals panel, which included Judges Karen Henderson, Gregory Katsas and Neomi Rao, pressed both sides on jurisdiction, procedure and whether the Pentagon had tried less intrusive options. Henderson called the DOD’s move a “spectacular overreach,” saying the record did not support a significant supply chain risk finding.

Anthropic sued Defense Secretary Pete Hegseth and the DOD in March after the agency said the startup posed a national-security risk. The designation requires defense contractors to certify they will not use Anthropic’s models in military work, even as the government continues using Claude for some operations.

• The DOD said Anthropic could encode limits into its models and interfere with military use.
• Anthropic said the blacklist is a contract dispute dressed up as a security action.
• The appeals court declined to block the designation in April, but fast-tracked the case.
• A separate San Francisco case gave Anthropic partial relief for non-DOD government users.

During arguments, DOJ lawyer Sharon Swingle said the designation was meant to alert the whole agency quickly so it could move to substitute AI systems. Anthropic lawyer Kelly Dunbar countered that the department is capable of procurement decisions without publicly branding an American company a security threat.

Why it matters

The dispute is about more than one vendor. If the Pentagon can blacklist a U.S. AI company during a contract fight, it could change how model providers negotiate guardrails for military use, especially around autonomous weapons and surveillance.

For developers and enterprise buyers, the case tests whether model restrictions written into training or deployment can be treated as a security hazard. It also shows how quickly AI revenue and valuation can keep rising even while public-sector access gets restricted.

Anthropic said last month it had reached $30 billion in annualized revenue, up from about $10 billion last year. CNBC also reported the company has been discussing a new round at a $900 billion valuation, far above its $380 billion valuation in February.

The court’s written opinion will decide whether the Pentagon’s label survives while the broader fight over AI use in defense contracts moves ahead. The bigger question: can a federal agency use a supply-chain risk label to settle a policy dispute with an AI vendor?