Sources: Pentagon Gives Anthropic an Ultimatum

Sources say Anthropic has been handed a blunt choice by the Pentagon: drop key AI ethics limits or risk being frozen out of federal work. Defense Secretary Pete Hegseth has given the San Francisco startup until 5:01pm Friday to accept the Trump administration's terms of use or face cancellation of its contract and designation as a "supply chain risk," according to an official and a person familiar with the talks who spoke to Politico. Axios, which first reported the news, notes defense officials are upset with Anthropic but also "worried about the consequences of losing access to its industry-leading model, Claude."

At a Tuesday meeting with Anthropic CEO Dario Amodei, Hegseth reportedly threatened to invoke the Defense Production Act, a Korean War-era law that lets the government direct private companies in wartime or emergencies. Amodei has drawn firm lines against using Claude for mass surveillance of Americans or for weapon targeting without a human in the loop—limits he reiterated in the meeting, the source says. A Pentagon official told Politico the clash is not about those issues and said the department is acting within the law. Anthropic, which confirmed the meeting, said discussions continue in "good faith" about how it can support national security while staying within what its models can "reliably and responsibly" do. How the Pentagon would both label Anthropic a risk and compel its cooperation remains unclear.