Now loading...
Anthropic’s CEO Dario Amodei is pushing for a resolution with the U.S. Department of Defense following a contentious fallout in talks that could sideline the AI firm from key military contracts. The renewed efforts come after negotiations crumbled last week, threatening to exclude Anthropic from the defense supply chain.
Amodei has engaged in direct conversations with Emil Michael, the undersecretary of defense for research and engineering, aiming to finalize terms that would grant the Pentagon ongoing access to Anthropic’s advanced AI systems. Sources familiar with the discussions say securing this agreement would allow the military to keep using the technology while averting a potential label as a supply chain threat—a step that Defense Secretary Pete Hegseth warned about on Friday but has not yet implemented.
The breakdown stemmed from irreconcilable differences over contract wording. Tensions boiled over when Michael publicly labeled Amodei a “liar” possessing a “God complex” during a Thursday exchange. By Friday, the two sides could not bridge gaps on provisions that Anthropic views as vital to blocking the use of its AI for widespread domestic surveillance—a core ethical boundary for the company, matched only by its opposition to lethal autonomous weapons.
In a candid internal memo to employees, first detailed by The Information and reviewed by journalists, Amodei recounted how the Defense Department proposed adopting Anthropic’s existing conditions but insisted on removing an important clause addressing “analysis of bulk acquired data.” He described this demand as deeply concerning, given its direct relevance to the surveillance risks the firm seeks to avoid. Amodei went further, accusing communications from both the Pentagon and OpenAI—which finalized its own deal with Hegseth on Friday—of spreading outright falsehoods or sowing deliberate confusion on these matters.
Amodei hinted at political motivations behind the impasse, noting that Anthropic’s reluctance to offer effusive endorsements of President Trump—unlike OpenAI’s Sam Altman—may have contributed to its isolation. The company initially secured a $200 million deal with the Defense Department in July of last year, marking it as the first AI provider to deploy models in classified environments and support national security operations.
Disputes intensified as the Pentagon advocated for broader permissions, insisting AI tools be permissible for all “lawful” applications. This push led Hegseth to announce plans for designating Anthropic a supply chain risk, which would force defense contractors to sever partnerships with the firm. The episode highlights growing frictions in how AI developers balance innovation with ethical safeguards in government collaborations.
Representatives from Anthropic and the Pentagon offered no immediate response to inquiries.
