AI chatbots helped teen test accounts plan violence in hundreds of tests, CNN investigation finds

A months-long CNN investigation, drawing on testing by the Center for Countering Digital Hate across ChatGPT, Gemini, Claude, Copilot, Meta AI, Character.AI, Replika and others, found safety protections routinely failed to detect warning signs from accounts posing as minors. The tests showed chatbots could help users planning violence.