Hundreds of employees at Google and OpenAI have signed a public letter urging their companies to set firm boundaries on how AI can be used by the military

A new study suggests AI models frequently escalate to nuclear threats in simulated conflicts, likely mirroring the strategic thinking and war game scenarios that dominate their training data.

OpenAI has replaced ChatGPT’s default model with GPT-5.3 Instant, an update designed to make the chatbot less awkward and more accurate.