Gain access to the White House Watch newsletter without cost
An essential overview of what Trump’s second term implies for Washington, commerce, and the globe
Anthropic’s CEO, Dario Amodei, is undertaking a final effort to secure an agreement with the U.S. defense department following the collapse of discussions last week, which jeopardized his company’s inclusion in the military’s procurement network.
Amodei has been engaged in talks with Emil Michael, the under-secretary of defense for research and engineering, aiming to finalize a contract that would regulate the Pentagon’s utilization of Anthropic’s AI models, as reported by several individuals familiar with the situation.
Securing a new accord would permit the U.S. armed forces to continue deploying Anthropic’s innovations and significantly mitigate the chance of the firm being labeled a supply chain risk — a measure warned by defense secretary Pete Hegseth on Friday but not yet implemented.
This endeavor to achieve a mutually acceptable pact follows the dramatic breakdown of talks last week.
Michael vehemently denounced Amodei as a “deceiver” afflicted with a “God complex” on Thursday.
Deliberations ceased a day later after the pair could not concur on phrasing that Anthropic deemed crucial for averting the use of AI for widespread internal monitoring, which represents one of the company’s firm boundaries, in addition to deadly autonomous armaments.
“Towards the close of negotiations, the [department] proposed to accept our existing conditions if we removed a particular clause about ‘analysis of bulk acquired data,’ which was the single line in the contract that precisely corresponded to this scenario we were most apprehensive of. We perceived that as highly dubious,” Amodei wrote in an internal communication to staff on Friday, which was initially disclosed by The Information on Wednesday and reviewed by the FT.
Within the memorandum, which is poised to hinder discussions, Amodei stated that much of the messaging from the Pentagon and OpenAI, which finalized its individual deal with Hegseth on Friday, comprised “absolute falsehoods about these matters or attempts to obscure the truth.”
Amodei posited that Anthropic had been excluded because “we haven’t offered autocratic adulation to Trump,” unlike OpenAI chief Sam Altman.
Anthropic was initially granted a $200 million contract with the U.S. defense department in July last year and was the pioneering AI system deployed in confidential environments and by national security organizations.
The dispute between Anthropic and the government intensified after the Pentagon advocated for AI companies to permit their systems’ deployment for any “permissible” objective.
This ultimately resulted in Hegseth announcing last week that he intended to categorize the company as a supply chain risk, thereby compelling firms in the military’s procurement network to sever relations with Anthropic.
Anthropic and the Pentagon chose not to provide a statement.
Further coverage from Joe Miller in Washington