Claude powers U.S. military Iran airstrike planning

Anthropic finds itself in an impossible position: its AI is being used for military targeting while it's suing the Defense Department over that very use — a tension that reveals how little control AI companies actually have once their models are deployed in enterprise pipelines.

• –Claude is integrated into Palantir's targeting software indirectly — Anthropic didn't sell directly to DoD, but the model still ends up in the kill chain
• –The Pentagon designated Anthropic a "national security threat" after the company tried to restrict military use for domestic surveillance and autonomous weapons — a designation that could boot it from all government contracts
• –Admiral Brad Cooper confirmed AI's role in target selection while insisting humans retain final authority — but "human in the loop" assurances are increasingly hard to verify in practice
• –Claude was also reportedly used in the U.S. operation targeting Venezuelan President Maduro, expanding the known scope of military AI use
• –This is a preview of the foundational AI governance question: who is liable when a model that "hallucinates" contributes to a targeting decision?