NatureAI models that lie, cheat and plot murder: how dangerous are LLMs really?Matthew Hutson2025-10-08Full textDOI https://doi.org/10.1038/d41586-025-03222-1