AI Willing to Use Nuclear Weapons in Wargames, Study Finds Amid Pentagon-AI Lab Standoff

A leading AI lab is facing pressure from the US military to hand over its technology, with concerns raised by a study showing AI models are willing to use nuclear weapons in wargames. The lab, Anthropic, is resisting unless the Pentagon agrees to their conditions, including no mass surveillance of civilians or lethal attacks without human oversight. The Pentagon has not revealed its plans for the AI technology, but is reportedly considering using Cold War-era laws to compel Anthropic to comply. The study by Professor Kenneth Payne found that AI models were prepared to use nuclear weapons in 95% of the wargames played, raising questions about AI safety versus supremacy.

Want More Context? 🔎