Anthropic Refuses Pentagon Demand to Remove AI Safety Guardrails, Risking $200M Contract

Summary

Anthropic has publicly refused the Pentagon's demand to remove safety restrictions from its Claude AI model, despite threats to cancel a $200 million contract and designate the company a "supply chain risk." Defense Secretary Pete Hegseth gave CEO Dario Amodei until Friday to grant the military unfettered access to Claude, specifically requesting the removal of guardrails preventing use in autonomous weapons systems and mass domestic surveillance. In a statement, Amodei said the company "cannot in good conscience" comply, arguing that such applications are "simply outside the bounds of what today's technology can safely and reliably do."

The standoff represents a high-profile test of Anthropic's positioning as the most safety-conscious major AI company. Until this week, Anthropic was the only AI provider approved for use in the military's classified systems, with its technology reportedly already deployed in military operations including the recent capture of Venezuelan leader Nicolás Maduro. The company received its DoD contract in July 2025 alongside other tech firms like Google and OpenAI.

Amodei emphasized Anthropic's desire to continue serving national security interests but only "with our two requested safeguards in place" against autonomous lethal weapons and mass surveillance. The company's refusal comes as the Department of Defense has increasingly integrated AI technology into military systems through lucrative contracts with major tech firms. Elon Musk's xAI recently reached an agreement to provide AI for classified systems, potentially offering the Pentagon an alternative provider.

• The standoff tests Anthropic's claims to prioritize AI safety over commercial interests, as the Pentagon increasingly integrates AI into military operations through contracts with major tech firms