A study finds GPT-5.2, Claude Sonnet 4, and Gemini 3 Flash deployed tactical nuclear weapons in 95% of 21 simulated war game scenarios, and never surrendered

In a study by Kenneth Payne at King’s College London, leading AI models from OpenAI, Anthropic, and Google recommended nuclear strikes in 95% of simulated war games. The AIs showed a willingness to use nuclear weapons without the same hesitation as humans, raising concerns about the potential impact of AI in real-world military decision-making.