Anthropic Rejects Proposal to Loosen AI Guardrails

Anthropic declines Pentagon’s request to relax AI safety, citing concerns over military applications like autonomous weapons.

Anthropic Rejects Pentagon Proposal to Loosen AI Guardrails

Anthropic has turned down a Pentagon proposal that would relax safety restrictions on its AI models, just before tomorrow’s deadline for an agreement or potential government action.

CEO Dario Amodei said the company cannot agree to military use of its technology without protections against applications like mass domestic surveillance or fully autonomous weapons, which it considers unsafe.

The Pentagon, led by Defense Secretary Pete Hegseth, has warned that failure to comply could lead to contract terminations, labeling Anthropic a supply chain risk, or invoking the Defense Production Act to compel cooperation.