ChatGPT can end conversations and pause accounts over a focus on violence

OpenAI has announced new safety measures in ChatGPT aimed at preventing violence. This is according to an analysis by Data Studios of OpenAI's announcement from October 30, 2026.

Stricter rules for teenagers

OpenAI already had a ban on helping teenagers obtain weapons. According to Data Studios, the ban is now being expanded to cover questions that appear innocent.

New rules for all users

For all users, ChatGPT can automatically end conversations in which the user appears to be fixated on violence or on seriously harming others.

After several violence-focused conversations, an account can also be temporarily paused. If earlier conversations have shown warning signs, ChatGPT will evaluate later requests more cautiously.

Information about past mass attacks will also be restricted. This applies to which weapons were used, how they were obtained, and comparisons with perpetrators.

Police referrals and upcoming report

OpenAI referred more than 200 cases to the police in the third quarter of 2026. It has not been disclosed how the cases are distributed across countries or what came of them.

OpenAI plans to publish a Community Safety Report starting in early 2027 and every six months thereafter. This is a plan and not something that has happened yet.

What has not been disclosed

According to Data Studios, OpenAI has not said when ChatGPT judges a conversation to be too focused on violence, or how accurately the system can recognize a teenager's age. Nothing has been disclosed about how to appeal an ended conversation or a paused account, either. The exact timeline and the regions where the measures apply are also not stated.

For ordinary users, this means ChatGPT may react more harshly if conversations turn to violence. How the rules work in practice is still unclear.