Defense Secretary Pete Hegseth has summoned Anthropic CEO Dario Amodei to the Pentagon on Tuesday for what sources describe as a tense showdown regarding military access to the company's Claude AI model.
"Anthropic knows this is not a get-to-know-you meeting," a senior Defense official told Axios. "This is not a friendly meeting. This is a sh*t-or-get-off-the-pot meeting."
At stake is the Pentagon's access to Claude, currently the only AI model approved for use on the military's classified systems. It is considered the most capable tool available for sensitive defense and intelligence operations. While the Department of Defense seeks to maintain this capability, officials are reportedly furious over Anthropic's refusal to fully lift its safety guardrails.
Negotiations appear to be nearing a breaking point. While an Anthropic spokesperson characterized the discussions as "productive" and "in good faith," Defense officials claim there has been zero progress.
Anthropic has expressed willingness to loosen usage restrictions but remains firm on walling off two specific areas: the mass surveillance of American citizens and the development of weapons systems that fire without human involvement. The spokesperson emphasized the company is "committed to using frontier AI in support of US national security."
Conversely, the Pentagon argues that the requirement to clear individual use cases with a private company is unduly restrictive. The department has demanded that all AI laboratories make their models available for "all lawful uses."
To force compliance, the Pentagon has threatened to designate Anthropic a "supply chain risk." Such a move would not only void existing contracts but also compel other defense contractors to certify they do not use Claude in their workflows, effectively blacklisting the company from the defense sector.
A Defense official indicated Hegseth intends to present Amodei with an ultimatum. However, officials acknowledge that offboarding Anthropic would be a massive logistical undertaking. The company is deeply entrenched in defense systems, and alternative AI labs currently offer capabilities viewed as inferior.
The dispute highlights a broader culture clash between Hegseth's Pentagon and the safety-focused Silicon Valley firm. Amodei has been vocal about the existential risks of AI, a stance the Pentagon views as obstructive.
"The problem with Dario is, with him, it's ideological," the senior Pentagon official said. "We know who we're dealing with."
Beyond the immediate confrontation, the meeting underscores unresolved questions regarding the role of AI in national security. Anthropic shares concerns with other experts that US law has not kept pace with the technology's ability to supercharge surveillance, alongside fears regarding the delegation of lethal force to autonomous systems.
Tensions between the two parties escalated in January following the reported use of Claude during a raid involving Venezuelan President Nicolás Maduro.
Joining Hegseth at the meeting will be Deputy Secretary Steve Feinberg and Under Secretary for Research and Engineering Emil Michael, who has led negotiations with Anthropic and three other model-makers. Anthropic declined to name the members of its delegation.