Anthropic Refuses DOD Request to Strip AI Safety Guardrails
Dario Amodei says Anthropic won't remove safeguards from its AI models at the Pentagon's request.
Anthropic is drawing a hard line against the Department of Defense. CEO Dario Amodei declared the company cannot "in good conscience" gut its AI safety measures at the DOD's request.
The stakes are real. Amodei said Anthropic will facilitate a smooth transition if the Pentagon decides to cut ties over the refusal. Getting dumped by the DOD is apparently preferable to shipping unguarded AI to the military.
Amodei framed national defense as existentially important, emphasizing his belief in leveraging AI to protect democracies and counter autocratic adversaries. But that commitment has limits — and those limits sit squarely at dismantling the safety infrastructure Anthropic has built around its models.
It's a rare move: a major AI company telling its biggest possible customer "no." Whether the Pentagon blinks or boots them remains to be seen.