The government relationship collapsed, according to Amodei, because they haven’t donated to or praised Trump in a dictator-like manner.
Anthropic CEO Dario Amodei is reportedly renegotiating with the Department of Defense, aiming to salvage ties with the US military and avoid being excluded as a “supply chain risk.” Previous discussions broke down after disagreements over granting Pentagon unrestricted AI access, with competitors like OpenAI stepping in.
Amodei is discussing a new contract with Under-Secretary Emil Michael to continue military use of Anthropic’s Claude AI models. This follows Michael’s public criticism of Amodei, accusing him of lying and risking national safety.
Securing a deal is crucial for Anthropic’s survival. Defense Secretary Pete Hegseth threatened to label the company a supply chain risk, typically reserved for foreign-linked threats, potentially isolating them within the tech sector.
A leaked memo from Amodei to staff further strains ties with the Trump administration, criticizing OpenAI’s Pentagon deal as misleading. He suggested their government relations soured due to not donating to Trump or offering excessive praise, contrasting with OpenAI’s Sam Altman, linked to Trump donations and connections.
Amodei claimed the Pentagon nearly agreed to their terms but wanted to remove a specific contract phrase, raising suspicions.
The dispute centers on the Pentagon’s call for full tech access and Anthropic’s refusal to cross red lines: no mass surveillance or lethal autonomous weapons. Despite pressure from Hegseth for “any lawful use,” Anthropic remains firm, unlike xAI and OpenAI.
Claude, once the sole AI with clearance for classified tasks, reportedly supported operations in Venezuela and Iran, influencing Trump’s reconsideration of the ban on federal use.