King's study finds AI chose nuclear signalling in 95% of simulated crises
AI Nuclear SigA study led by Professor Kenneth Payne examined how AI models handle simulated nuclear crises. Three AI models, GPT-5.2, Claude Sonnet 4, and Gemini 3 Flash, engaged in 21 nuclear crisis scenarios, generating over 780,000 words of reasoning. The models predominantly used nuclear signaling, with 95% of scenarios involving mutual signaling, though actual nuclear escalation was rare. The study revealed AI's tendency to use nuclear weapons as compellence tools rather than deterrence. It also highlighted how deadlines significantly influenced AI escalation behavior.
Show original excerpt (English · first 3 paragraphs)
The study, led by Professor Kenneth Payne from the Department of Defence Studies, examined how large language models (LLMs) navigate simulated nuclear crises. As militaries and security institutions increasingly experiment with AI-assisted analysis and wargaming, understanding how such systems reason under pressure is becoming increasingly critical.
Three leading AI models – GPT-5.2, Claude Sonnet 4 and Gemini 3 Flash – were placed in a tournament of 21 simulated nuclear crisis scenarios. Across 329 turns of play, the models generated approximately 780,000 words of structured reasoning – more than the combined length of War and Peace and The Iliad.
All 21 crisis games featured nuclear signaling by at least one side, and 95% involved mutual nuclear signaling. However, while models readily threatened nuclear action, crossing the tactical nuclear threshold was less common, and ‘strategic’ full-scale nuclear war was rare.
* For copyright reasons we quote only the first 3 paragraphs. Read the full article at the source.