Anthropic vs. Pentagon: AI Ethics Clash Over Claude AI Use

Emails between Anthropic CEO Dario Amodei and a Pentagon official reveal how a $200 million relationship soured. The unsealed court documents show negotiations over the military's use of Anthropic's AI model, Claude. Amodei set two conditions: no autonomous weapons and no domestic mass surveillance. He stated these boundaries would prevent using powerful AI to spy on Americans while allowing foreign intelligence work. The Pentagon, however, wanted broader terms, insisting on "all lawful uses." A Pentagon official called Anthropic's proposal "not workable," saying it clashed with their principles. The disagreement centered on the distinction between defensive and offensive weapons. The Pentagon argued that a drone capable of defensive actions could also be used offensively. Amodei countered that following existing law was insufficient because US law permits some domestic surveillance. This breakdown led to Anthropic being labeled a "supply-chain risk," a designation usually reserved for foreign adversaries. This situation highlights the ongoing tension between AI ethics and military application.