AI models that lie, cheat and plot murder: how dangerous are LLMs really? | Not Hacker News!