Bold claim: Anthropic refuses the Pentagon’s demand, insisting their concerns about AI being used for mass surveillance or in autonomous weapons cannot be ignored. And this is where the controversy starts to unfold...
Anthropic has rejected the latest language from the Pentagon intended to modify their contract. The company argues that the proposed changes fail to address core worries about how Claude—the first AI system deployed within the military’s classified network—could be misused.
The dispute centers on restrictions governing Claude’s use. Defense Secretary Pete Hegseth reportedly told Anthropic CEO Dario Amodei that if Anthropic won’t permit Claude to be used for all lawful purposes, the Pentagon would cancel the $200 million contract. He also suggested that failure to comply could label Anthropic a “supply chain risk,” a designation typically reserved for firms tied to foreign adversaries.
Anthropic contests the Pentagon’s new wording, describing it as a compromise that is paired with legal language allowing safeguards to be bypassed at will.
In a detailed blog post published Thursday, Amodei stated that AI has existential importance for defending democracies, including the United States, and for countering autocratic rivals. He acknowledged that military decisions lie with the Pentagon, not private companies, but warned that AI could in a narrow set of scenarios either undermine or threaten democratic values. He cited mass surveillance and autonomous weapons as examples of use cases that current technology cannot safely and reliably handle.
Amodei reiterated that the Pentagon’s pressure does not alter Anthropic’s stance: they cannot, in good conscience, agree to the proposed terms.
There was no immediate comment from the Pentagon when contacted for a response.
