OpenAI unveils new safety rules for teen AI use

OpenAI has published an update to its Model Spec with new Under-18 (U18) Principles designed to make ChatGPT safer and more appropriate for teens aged 13 to 17. Developed with input from experts like the American Psychological Association, these rules are grounded in developmental science and prioritize prevention, transparency, and early intervention. The guidelines put teen safety first—even when it conflicts with other goals—encourage real-world support like offline relationships, treat teens as teens without condescension or treating them like adults, and emphasize clear expectations. In practice, this means stronger guardrails for topics like self-harm, romantic roleplay, graphic content, dangerous activities, body image issues, and requests to keep secrets. The model is also trained to guide teens toward trusted offline resources when safety concerns arise. Why it matters: As AI becomes a daily tool for younger users, this update marks a shift from one-size-fits-all moderation to developmentally-aware AI design, but parents and schools still need to stay actively involved in how teens use these systems.