Pentagon official says Anthropic negotiations are ‘over’ as supply-chain dispute heads to court

A senior Pentagon official has all but closed the door on reviving negotiations with Anthropic over the military use of the company’s artificial intelligence tools.

Emil Michael, the Under Secretary of Defence for Research and Engineering, told Bloomberg that Anthropic’s decision to sue the department was “an expected response” and that the company’s legal effort to reverse its supply-chain exclusion would not alter the Pentagon’s position. “I don’t think there’s a scenario where this resolves that way,” Michael said in the interview.

Anthropic filed suit to prevent the Pentagon and other federal agencies from designating it a threat to the US supply chain and barring it from government contracts. In court filings submitted Monday, the company argued that those measures violate its rights to free expression and due process under the US Constitution. An Anthropic spokesperson declined to comment on Michael’s remarks.

Michael — a former Uber executive who now oversees the Pentagon’s efforts to accelerate AI adoption — had spent weeks in tense negotiations with Anthropic chief executive Dario Amodei over the terms of use for the company’s AI tools. Those talks collapsed approximately two weeks ago after Anthropic sought assurances that its AI would not be deployed for the mass surveillance of Americans or for the operation of autonomous weapons. The impasse prompted the Pentagon to declare the San Francisco-based company a supply-chain risk — a designation that is ordinarily reserved for firms headquartered in adversarial nations.

Until recently, Anthropic was the sole provider of an AI system capable of operating within the Pentagon’s classified cloud environment, and its Claude model had become the preferred tool among defence personnel, valued for its ease of use. The designation now jeopardises a $200 million contract under which Anthropic was supplying classified AI capabilities to the Pentagon, and could prevent the company from partnering with other firms on military work.

Defence Secretary Pete Hegseth and President Donald Trump have set a six-month deadline for the military and other US agencies to migrate from Anthropic’s systems to alternative AI providers.

In one of its court filings, Anthropic disclosed concern that the government’s actions are already reverberating through its broader federal contracting relationships. The company said a vendor with which it collaborates on specialised applications had warned it might “pause work or even remove Claude from existing applications.” Anthropic further noted that other contractors had “expressed concerns, paused collaborations, and indicated they are considering terminating contracts.”

Michael reiterated a central Pentagon grievance: that Anthropic’s publicly stated stance on safeguards for military applications signalled a desire to exercise authority over the chain of command and, potentially, over operational decisions. Anthropic has rejected that characterisation, stressing that all military determinations rest exclusively with the Pentagon.

Amodei has said the company wishes to continue working with the military, but is seeking the incorporation of the two usage restrictions into all defence contracts. When asked whether negotiations could resume, Michael was unambiguous: “Negotiations are over. We’re moving on.”