OpenAI tightens ChatGPT rules on violence and teen weapons questions

OpenAI said on Oct. 30 that, over the next few weeks, ChatGPT will give all users less detail on past mass attacks, including weapons and how they were obtained. It will also limit practical weapons and security-system information for people predicted to be under 18, and it may end a chat or pause an account when recent conversations show fixation on seriously harming someone. The company said it made more than 200 law-enforcement referrals in the third quarter of 2026 and will publish a first Community Safety Report in early 2027.