OpenAI will publish a Community Safety Report and tightens ChatGPT limits on violence and weapons for teens

OpenAI announced on 30 October that it will start publishing a Community Safety Report and that it is tightening how ChatGPT handles violence and weapons, with extra limits for teens. The changes roll out over the next few weeks.

The report

OpenAI says it made more than 200 referrals to law enforcement in the third quarter of 2026. The first Community Safety Report, covering the second half of 2026, is due in early 2027, with new ones every six months after that.

Changes for everyone

• Past attacks: ChatGPT will restrict detailed reconstructions of past mass attacks, including the weapons and equipment used, how they were obtained and comparisons of scale.
• Warning signs: when recent conversations show warning signs, ChatGPT will answer more conservatively. It can end a chat that keeps returning to violence and may temporarily pause an account.

Changes for teens

Teen accounts get limits on practical information about weapons and security systems even when no harmful intent is shown. OpenAI says it is improving age prediction, and adults misclassified as teens can verify their age with Persona. It is also working on notifying parents when a linked teen account is paused.

Why this matters

A regular report with counts gives outsiders some way to see how often ChatGPT’s safety systems lead to real-world action, which until now has only come out in individual cases. The new limits will also reach ordinary users: writers, researchers and journalists covering past attacks may see more refusals, and age prediction errors will land on some adults. It fits a wider push on minors’ safety, such as New York’s ban on unsafe companion features, and follows OpenAI’s updated Model Spec a day earlier.