TL;DR
Anthropic has refused a Pentagon ultimatum to remove safety restrictions from its Claude AI model, saying it “cannot in good conscience” allow the technology to be used for mass domestic surveillance or autonomous weapons systems that can kill without human input. The Defence Department threatened to cancel a £148 million contract and designate Anthropic a “supply chain risk.”
The standoff
Defence Secretary Pete Hegseth gave Anthropic CEO Dario Amodei until Friday to agree to allow any lawful use of Claude without restrictions. Anthropic pushed back on two specific applications: mass domestic surveillance and autonomous weapons that operate without human oversight.
“Using AI for autonomous weapons and mass domestic surveillance is simply outside the bounds of what today’s technology can safely and reliably do,” Amodei said. He added that the company’s “strong preference is to continue to serve the Department and our warfighters — with our two requested safeguards in place.”
What’s at stake
Anthropic was the only AI model approved for use in the US military’s classified systems until this week, when Elon Musk’s xAI reached a similar agreement. The company’s technology has reportedly already been used in military operations, including last month’s capture of Venezuelan leader Nicolás Maduro.
If Hegseth follows through, a “supply chain risk” designation — typically reserved for foreign adversaries — would prohibit other military vendors from using Anthropic’s products. The financial impact would extend well beyond the £148 million contract.
Wider context
The dispute comes as Anthropic this week also watered down a core policy against releasing new AI models without first guaranteeing their safety. The company has long positioned itself as the industry’s most prominent advocate for AI regulation and safety precautions, even while expanding military partnerships.
Looking forward
Whether Anthropic holds its position will test the limits of corporate safety commitments when they conflict with government demands. The outcome could set a precedent for how other AI companies negotiate military contracts and where they draw lines on AI use.