Anthropic suffers setback in Pentagon supply-chain risk fight

Anthropic just hit a major legal roadblock in Washington, even after winning a temporary reprieve on the West Coast.

A federal appeals court in D.C. on Wednesday denied Anthropic's bid to pause the Pentagon's decision designating the AI company as a supply-chain risk.

Anthropic had pushed for an emergency stay to head off financial hit and reputational damage from the Pentagon blacklist.

A tale of two courtrooms

The dispute centers on Department of Defense restrictions against deploying Anthropic's Claude models inside classified environments.

Last month, Anthropic scored a win when a San Francisco federal judge granted a preliminary injunction preventing the broader administration from banning Claude across government agencies.

Because of that West Coast ruling, civilian and non-defense agencies are not forced to cancel their existing contracts with the AI firm.

Wednesday's D.C. ruling throws a wrench into that momentum. According to a source familiar with the case, the Pentagon can still exclude Anthropic from any new military contracts under a separate statute.

The same source noted that the kind of emergency relief Anthropic sought from the D.C. appeals court is rarely granted in practice.

Interestingly, the Pentagon isn't pulling the plug immediately. The agency is still expected to keep using Anthropic's tools for the next six months.

Anthropic responded that it remains confident the courts will ultimately rule the Pentagon's supply-chain designations unlawful, adding that the litigation was necessary to protect its partners and customers.

Why it matters

For enterprise customers and developers building on Claude, there is no immediate reason to switch models. Non-defense agencies are keeping their contracts, and even the Pentagon plans to stick with Anthropic products for another six months.

However, getting boxed out of brand-new military contracts hurts Anthropic where it stings most. Government and defense spending represents billions in potential revenue, and rivals are eager to fill that void.

We think Anthropic was right to take a combative stance in court. Allowing a supply-chain risk label to stand unanswered would give risk-averse enterprise buyers cold feet well beyond the military sector.

The paradox is clear: the Pentagon still relies on Claude enough to keep running it for half a year, yet legally bars it from bidding on what comes next. How this split legal fight concludes will set the rules of engagement between frontier AI labs and national security agencies for years to come.