TL;DR
- A federal judge told the courtroom the Pentagon’s actions against Anthropic “look like an attempt to cripple” the company
- Anthropic is seeking an injunction against the Department of Defense’s unprecedented designation of a US company as a supply chain risk
- The case stems from Anthropic’s refusal to allow its Claude AI to be used for domestic mass surveillance or fully autonomous lethal weapons
Safety stance meets government retaliation
Anthropic and the US Department of Defense appeared in a northern California district court on Tuesday in what may become a defining case for AI safety policy. Judge Rita Lin heard arguments on Anthropic’s request for a temporary injunction against Defence Secretary Pete Hegseth’s decision to classify the company as a supply chain risk — the first time this designation has been applied to a US firm.
The dispute centres on Anthropic’s refusal to loosen safety restrictions on Claude, its AI assistant, for military applications including mass domestic surveillance and fully autonomous weapons systems. The company argues its technology is not reliable enough for such uses, a position its chief executive Dario Amodei has framed in terms of preventing authoritarian applications of AI.
A judge’s pointed questioning
Judge Lin described the case as a “fascinating public policy debate” but made clear her role was limited to assessing legality. Her questioning of government lawyers was notably direct. When the defence department’s legal team argued that Hegseth’s social media post barring contractors from working with Anthropic had no legal force, Lin pressed the point: “You’re standing here saying, ‘We said it, but we didn’t really mean it.’”
The government’s lawyer conceded he could not explain why Hegseth would issue the directive if it carried no legal effect.
Deep entanglement complicates separation
Unpicking Claude from federal operations would be a substantial undertaking. The technology is reportedly embedded across multiple government agencies, including military operations where it has been used for target selection and analysis during strikes in Iran. Anthropic argues the supply chain designation alone threatens hundreds of millions in revenue and would cause irreparable commercial harm.
Meanwhile, the Pentagon has signed new deals with OpenAI and Elon Musk’s xAI for classified environment access, signalling a willingness to replace Anthropic with competitors less resistant to military applications.
What UK observers should note
For Britain’s AI sector, this case carries implications beyond US domestic politics. The UK government has positioned itself as a partner to both the US defence establishment and the AI safety community. A precedent where an AI company faces punitive government action for maintaining safety boundaries would complicate the UK’s own approach to AI governance, particularly as British firms weigh the commercial risks of principled safety stances.
The outcome will also matter for UK defence procurement. If US-based AI providers face pressure to choose between safety commitments and government contracts, allied nations relying on the same tools will need to consider what compromises are baked into the products they adopt.
Judge Lin has not yet issued a ruling.