OpenAI Limits Weapons Information for Under-18 Users, Plans to End Chats That Show Fixation on Violence

OpenAI announced on October 30 (local time) that it is strengthening ChatGPT’s violence prevention measures. Under the new approach, users under 18 will face tighter limits on weapons-related information, and the system will step in earlier when a conversation shows signs of violence risk. OpenAI said it will begin rolling out the safeguards it described over the coming weeks.

Which questions get blocked

Start with the restrictions that apply to every user. Content that recreates past mass attacks in detail will be blocked. That includes descriptions of the weapons and equipment used in an attack, and of how the attacker obtained them at the time. Detailed comparisons of the scale of harm, which can amplify an attacker’s notoriety, are also restricted. These limits apply regardless of a user’s age or stated reason for asking.

OpenAI is also developing an approach that looks beyond a single question to warning signs that have appeared across a recent conversation. If there are signals suggesting the potential for violence, the model will be more cautious about sharing information that could be misused. The company is also moving toward automatically ending conversations when there are signs of a fixation on violence or on causing serious harm to others. If such terminations happen repeatedly, OpenAI is considering temporarily suspending the account.

What changes for teen accounts

Users under 18 will face additional restrictions even when no violent intent is apparent. The restrictions target information about weapons and security systems that could realistically be misused. Conversations aimed at preventing violence, de-escalating a situation, or seeking help remain available.

Age determination does not end at sign-up. OpenAI said it will continue to examine a range of signals, including usage patterns, and will move users into the under-18 experience when warranted. Adults who are misclassified as teens can verify their age through OpenAI’s partner Persona and have the extra protections lifted.

Parent notifications are already in place. When a teen account is deactivated for violating violence-related policies, a notification goes to the linked parent account. A feature that notifies parents when an account is temporarily suspended after repeated violence-related conversations is still in development.

What goes into the safety reports

OpenAI plans to publish community safety reports on a regular basis. The first report will cover the second half of 2026 and arrive in early 2027, with updates every six months after that. The reports are set to cover how the company detects, reviews, and responds to violence risk, along with de-identified case examples, countermeasures and policy changes, and related data. OpenAI said it referred more than 200 cases to law enforcement in the third quarter of 2026.

The center of gravity in these measures has shifted from simply blocking answers to intervening based on the flow of a conversation and the user’s age. For users, that means both the range of information they can get and whether they can keep a conversation going are changing at the same time. The thing to watch in the reports to come is how OpenAI actually handled real risks, and how that experience feeds back into policy.