OpenAI tightens violence safeguards in ChatGPT and adds teen protections, plans a Community Safety Report
OpenAI announced expanded violence-prevention safeguards in ChatGPT, tighter weapons-related limits for teens, and a new recurring Community Safety Report. The safeguards begin rolling out over the next few weeks.
Details
- Law enforcement referrals: OpenAI says it made over 200 referrals to law enforcement in Q3 2026 for imminent threats of violence, relying on mental health and behavioral experts for difficult cases.
- Community Safety Report: The first report is planned for early 2027 and will cover the second half of 2026, with updates every six months. It will describe how systems are trained to detect, investigate and respond, with anonymized examples, mitigations, policy changes and supporting data.
- Limits for all users: ChatGPT will give less detailed information about past mass attacks and perpetrators, including reconstructions (weapons, equipment, how they were obtained) and scale-of-harm comparisons that could raise an attacker’s notoriety. Users can still learn about the events, their impact and warning signs. This applies regardless of age or stated reason.
- Warning signs: Building on work from May, ChatGPT responds more conservatively when recent chats show possible signs of violence. Accounts violating violence policies are banned. OpenAI is developing automated early intervention, so ChatGPT can end chats showing fixation on violence; repeated cases may lead to a temporary account pause.
- Teens: Under ChatGPT for Teens, weapons-related information is further limited, including practical information about weapons and security systems. Teens can still get help de-escalating, finding support and reaching a trusted adult.
- Age prediction: The system now assesses signals continually and moves likely teens into the under-18 experience. Adults placed there by mistake can verify their age through Persona.
- Parental controls: Linked parents get alerts for serious self-harm concerns or disordered-eating behaviors. Since July, alerts also cover a teen account deactivated for violence-policy violations; OpenAI is working on alerts for accounts paused for repeated violent discussion.
Outlook
Some pieces (automated chat ending, pause notices to parents) are still in development, and OpenAI gives no exact rollout dates beyond the next few weeks.