Anthropic sues US Defense Department over supply chain risk label

Anthropic sued the US Department of Defense in California and filed a petition in Washington, DC to contest a DOD decision naming it a national security supply chain risk after a dispute over Pentagon access to its AI.

Anthropic said it had two “red lines”: it did not want its models used for mass surveillance of Americans and said its models were not ready to power fully autonomous weapons.

Pete Hegseth, defense secretary, said the Pentagon should have access to AI for “any lawful purpose” and not be limited by contractors.

The designation requires companies and agencies that work with the Pentagon to certify they do not use Anthropic models, and, following public statements by President Trump and Hegseth, the GSA ended Anthropic’s OneGov contract.

In its San Francisco lawsuit, Anthropic called the designation “unprecedented and unlawful” and accused the administration of retaliating against its protected speech on AI safety.

Anthropic asked a San Francisco federal court to pause and block the designation and filed a separate petition to the DC Circuit under federal procurement law.

🔗 Source: TechCrunch

🧠 Food for thought

Implications, context, and why it matters.

The lawsuit follows the abrupt end of a deep government partnership

Before the dispute, Anthropic was the first AI frontier lab cleared for use by U.S. officials on classified networks 1.
The now-ended OneGov agreement aimed to speed up AI use by offering Anthropic’s Claude products to all three branches of the federal government for $1 2.
The “supply chain risk” label is a tough measure usually saved for foreign adversaries, and applying it to a U.S. company is widely seen as unusual 3.
The label works like a blacklist that stops Pentagon suppliers from using Anthropic’s Claude. The Trump administration has framed it to require defense contractors to certify they do not use Claude on work tied to the Department of Defense 1.

The Pentagon’s ban immediately reshapes the government AI market

Anthropic’s blacklisting opens room for rivals, with OpenAI announcing a Pentagon deal hours after the government penalized Anthropic over its position 3.
The Pentagon is exploring moving work previously done on Claude to models from Google, OpenAI, and Elon Musk’s xAI (his AI company) 3.
The standoff leaves AI vendors weighing ethics rules for military use against the risk of losing access to government contracts 4.
Using a supply-chain risk label to resolve a policy fight sets a precedent, which could shift how U.S. tech companies handle sensitive government partnerships 5.