New court filing reveals Pentagon told Anthropic the two sides were nearly aligned — a week after Trump declared the relationship kaput

In a recent legal development, Anthropic has filed sworn declarations in a California federal court, challenging the Pentagon's assertion that the AI firm presents an 'unacceptable risk to national security.' Anthropic contends that the government’s position is based on misunderstandings and allegations that were never addressed during prior negotiations. These declarations accompany Anthropic’s response in its ongoing lawsuit against the Department of Defense, with a crucial hearing scheduled for March 24 before Judge Rita Lin in San Francisco. The conflict originated in late February when President Trump and Defense Secretary Pete Hegseth announced a severance of ties with Anthropic, following the company's refusal to permit unrestricted military applications of its AI technology. The declarations were authored by Sarah Heck, Anthropic’s Head of Policy, and Thiyagu Ramasamy, the Head of Public Sector. Heck, who previously served on the National Security Council under the Obama administration, attended a pivotal meeting on February 24, where CEO Dario Amodei discussed the situation with Defense Secretary Hegseth and Under Secretary Emil Michael. In her statement, Heck refutes claims that Anthropic sought a role in military operations, asserting, 'At no time during Anthropic’s negotiations did I or any other employee demand such authority.' She highlights that concerns about the possibility of Anthropic disabling its technology were not raised during negotiations. Moreover, Heck's declaration mentions a March 4 email from Under Secretary Michael, indicating that the two parties were 'very close' on issues surrounding autonomous weapons and surveillance, the very points now cited as evidence of Anthropic’s alleged threat to national security. This email is particularly noteworthy against the backdrop of the subsequent public statements from Pentagon officials, which contradicted the notion of ongoing productive discussions. Ramasamy, who previously managed AI projects for government clients at Amazon Web Services, counters the Pentagon's claims regarding the potential for Anthropic to interfere with military operations. He emphasizes that once the Claude model is integrated into a secure, air-gapped system, Anthropic has no access to it and cannot implement unauthorized changes. He argues that any alteration would necessitate explicit approval from the Pentagon. Furthermore, Ramasamy addresses concerns over the hiring of foreign nationals, asserting that Anthropic employees have undergone rigorous U.S. government security clearance processes. He notes that Anthropic is likely the only AI company where cleared personnel were directly involved in creating AI models intended for classified settings. In its lawsuit, Anthropic argues that the supply-chain risk designation is an act of government retaliation for its publicly expressed views on AI safety, which they contend violates the First Amendment. The government's response, a 40-page document, dismissed this interpretation, framing Anthropic’s refusal to allow military use of its technology as a business decision rather than a matter of protected speech, asserting that the designation was strictly a national security measure and not punitive in nature.