High-risk autonomous behaviours are an increasingly prevalent and dangerous reality for frontier AI models www.wired.com/story/openai...

• wired.com / story/openai-models-escaped-containment-and-hacked-huggingface