Dario Amodei publicly refuses Pentagon demand for 'unfettered' Claude access

Dario Amodei published a statement describing the state of Anthropic’s contract negotiations with the US Department of War, and refusing a demand he said the department had made: that Anthropic accept “any lawful use” of Claude by the military and drop two safeguards the company treated as non-negotiable. The first barred Claude from powering fully autonomous weapons systems without human decision-making in the loop; the second barred its use for mass domestic surveillance of Americans.

Amodei argued the autonomous-weapons restriction rested on a capability limit rather than a moral objection alone — “frontier AI systems are simply not reliable enough to power fully autonomous weapons,” he wrote, and the oversight structures that would make such use safe do not yet exist. On surveillance, he argued AI could turn scattered personal data into comprehensive profiles of individuals “automatically and at massive scale,” a capability he said posed a novel threat to civil liberties that existing law had not anticipated. He described the department’s position as contradictory: threatening to bar Anthropic from government business as a security risk while simultaneously treating Claude as essential to national-security work. Anthropic said it would help ensure a smooth transition if the department chose another provider rather than accept the terms.

The statement was the opening move in a dispute that escalated rapidly: within a day, President Trump ordered federal agencies to stop using Anthropic’s technology, and Defense Secretary Pete Hegseth said he would designate the company a “supply-chain risk” — a move Anthropic disputed had actually taken formal effect. The confrontation set the terms for a legal and political fight over how far a vendor could bind its own product against military use that continued for months.