Anthropic Fights Pentagon Risk Designation; CEO Apologizes

Amodei stated the designation, received March 4th, is "not legally sound" and misapplies statute 10 USC 3252, which he describes as narrow and intended to protect the government, not punish suppliers. Concurrently, Amodei apologized for a leaked internal memo criticizing President Donald Trump, clarifying it was an "out-of-date assessment" written during a "difficult day" following confusing announcements regarding Anthropic and OpenAI.

Anthropic emphasizes the SCR designation's limited scope, asserting it only affects the use of its Claude AI within direct DoW contracts, leaving commercial and individual customer access completely unaffected. The company reiterates its commitment to US national security, offering its models to the DoW at nominal cost, with specific exceptions for autonomous weapons and mass domestic surveillance.

This situation highlights growing tensions and regulatory challenges for AI firms operating in sensitive government sectors.