Can AI firms set limits on the military’s use of their models? Anthropic is in intense talks with the Pentagon after rejecting military contract terms that require loosening restrictions on its AI models, allowing “any lawful use,” including mass surveillance of Americans and fully autonomous lethal weapons.
Pentagon CTO Emil Michael may label Anthropic a “supply chain risk” if it doesn’t comply; this label is mostly for national security threats. OpenAI and xAI reportedly agreed to the terms, but after a White House meeting with Defense Secretary Pete Hegseth, Anthropic CEO Dario Amodei is still refusing, stating “threats do not change our position: we cannot in good conscience accede to their request.”
As Anthropic faces a Pentagon ultimatum to allow US military access to its technology, tech workers ponder the implications of their companies’ government contracts. The Department of Defense has pushed Anthropic to remove its AI guardrails, allowing military use with no human oversight. OpenAI and xAI have reportedly accepted these terms, though OpenAI seeks similar red lines to Anthropic. Workers feel betrayed; as one Amazon Web Services employee says, “I thought tech was about making people’s lives easier, but now it seems like it’s all about surveillance and harm.”
Less than 24 hours before the Pentagon deadline, Anthropic rejected demands for unrestricted AI access. This stems from public statements and negotiations, where Defense Secretary Pete Hegseth seeks renegotiations of AI contracts. Anthropic insists on no mass surveillance or lethal autonomous weapons, unlike OpenAI and xAI, which reportedly agreed to new terms. Amodei was summoned to the White House, where Hegseth reportedly demanded compliance by end of the workday on Friday.
Anthropic’s battle with the Department of Defense has unfolded via social media and public statements. The key point is the term “any lawful use,” giving the US military rights for mass surveillance and autonomous lethal AI. Negotiations have soured, with CTO Emil Michael pushing for Anthropic to be labeled a “supply chain risk.” This label is for national security threats, like cyber warfare. Amodei is expected to meet Hegseth at the Pentagon in a “make or break” meeting.