A new study at King’s College London, found that leading ChatGPT, Claude, and Gemini deployed the use of nuclear weapons in nearly all simulated war game scenarios designed to test strategic escalation. In simulations of armed conflict, AI consistently chose tactical or strategic nuclear deployment and never opted for accommodation or surrender, even after models were reminded of the devastating consequences of nuclear war. Researchers say the results highlight potential risks if AI systems are ever used in high-stakes military decision-making, underscoring that current models may not interpret stakes or humanitarian costs the way humans do. Source: NewScientist Follow futurewalt.ai for latest updates in AI industry #nuclear #ai #aiwar