OpenAI tightens ChatGPT safety rules and will release safety reports

OpenAI has announced new safety upgrades for ChatGPT. It will release a Community Safety Report regularly, with the first one coming in early 2027 and updates every six months. In just one quarter, OpenAI passed over 200 urgent violence threat cases to law enforcement. For past mass attacks, ChatGPT will now hold back details like weapon specs, how to get them, and harm comparisons, no matter who asks. But people can still learn about the events, the victims, the impact, and how to spot warning signs. If a chat shows someone is obsessed with violence, ChatGPT can end it right away, and repeated breaks may lead to a temporary account freeze. Rules for teens on gun and weapon info are stricter too, and a feature to alert parents is in the works.