The Pentagon and AI leader Anthropic are at an impasse over military access to Claude, its newly developed artificial intelligence model, sources say. Defense Secretary Pete Hegseth met with Anthropic’s CEO Dario Amodei Tuesday morning in a high-stakes discussion about the use of Claude for military operations.
Hegseth demanded that Anthropic submit a signed document granting full access to Claude by this Friday or face potential legal action under the Defense Production Act, sources told CBS News. The Pentagon awarded Anthropic a $200 million contract in July to bolster its AI capabilities for national security.
Amodei and his team have repeatedly sought safeguards from military use of Claude, stressing concerns about mass surveillance and autonomous weapons, among other issues. Officials argue that the Defense Department’s lawful orders suffice without human intervention or restriction.
Anthropic also demands that Claude operates under strict legal confines, ensuring its use aligns with ethical standards in warfare. The company has consistently pushed for measures to prevent potentially lethal mistakes resulting from the AI’s unreliability.
In response, Hegseth invoked an analogy of government contracts for hardware like aircraft—where suppliers have no say over how customers might use their products. He contended that Claude should be treated similarly, asserting legal authority to dictate its deployment within military operations.
The meeting escalated after Amodei’s departure with discussions revolving around the invocation of the Defense Production Act. However, given the Pentagon’s skepticism regarding Anthropic’s trustworthiness at present, they are considering other measures such as designating Anthropic as a “supply chain risk.”
Despite these tensions, an Anthropic spokesperson has stated that negotiations continue in good faith to ensure continued government support for national security missions. Hegseth set a deadline of 5 p.m. Friday.
This standoff underscores the complex intersection between military necessity and ethical concerns over AI technology’s utilization. The stakes are high as both parties seek control and oversight, while also grappling with potential unintended consequences and limitations in AI capabilities.