Technology · 27 February 2026
AI chose nuclear war 95% of time. Now a Pentagon standoff.
A study by Professor Kenneth Payne at King's College London found that three leading AI models — OpenAI's GPT-5.2, Anthropic's Claude Sonnet 4 and Google's Gemini 3 Flash — deployed nuclear weapons in 95% of 21 simulated war games, with no model ever choosing to surrender. The findings were published as US Defence Secretary Pete Hegseth gave Anthropic a Friday deadline to grant the Pentagon unrestricted military access to Claude, threatening the company's government contracts if it refused. Anthropic's CEO Dario Amodei said the company could not comply with the request in good conscience.
Reported by Sky News · How we write briefs