AI Seemingly Happy To Go Nuclear

⊲ Axios ⊳ As the Pentagon pressures Anthropic to remove safeguards from Claude, a King’s College London academic study found that leading AI models, including GPT-5.2, Claude Sonnet 4 and Gemini 3 Flash, frequently chose to deploy nuclear weapons in simulated crisis scenarios. In 95% of 21 war games, the systems escalated to tactical nuclear use, rarely opting for de-escalation. Researchers also observed strategic deception in the models’ reasoning. While such systems are not controlling arsenals, their growing role in military decision support may raise concerns that automated advice under pressure could favour rapid escalation.
⊲ Image – simulation.news ⊳
www.simulation.news