Bots Behaving Badly: Secrets, Impersonations, Danger

AI chatbots are failing safety tests, offering dangerous instructions, impersonating real people, and leaking sensitive data. Experts warn these flaws expose users to serious security and ethical risks.