THE ESSENTIALS IN 30 SECONDS

  • Users identified as under 18 will get less practical information about weapons and security systems, even when a question shows no violent intent.
  • ChatGPT will be able to end conversations that show a fixation on violence, and repeated cases may trigger a temporary account pause.
  • OpenAI reports more than 200 referrals to law enforcement in the third quarter of 2026; the safeguards start rolling out over the next few weeks.

OpenAI announced on October 30 a set of ChatGPT changes aimed at preventing real-world violence, with the strictest rules reserved for teenagers. According to Reuters, the company says the safeguards will begin rolling out over the next few weeks and will be refined over time.

Stricter limits for under-18 users

For accounts placed in the teen experience, ChatGPT will hold back practical details about weapons and about security systems that could be misused. OpenAI says these limits apply even when a question looks harmless and the conversation contains no sign of violent intent. Teens can still get help with de-escalation, violence prevention and reaching a trusted adult, Unite.AI reports.

The company also says it has strengthened its age prediction system, which estimates from behavioral and account signals whether a user is likely under 18 and moves them into the teen version. Adults placed there by mistake can confirm their age through Persona, OpenAI's verification partner. OpenAI already alerts parents when a teen's account is deactivated for violating its violence policies, and says it is working on notifications for temporary pauses as well.

Changes for all users

Some measures apply regardless of age. ChatGPT will give less detail about past mass attacks and their perpetrators, including the weapons used and how they were obtained, even when that material is available elsewhere online. Questions about the impact of such events, warning signs and prevention remain possible.

OpenAI is also testing automated interventions: ChatGPT can end a conversation in which someone appears fixated on violence or on seriously harming another person, and several ended conversations may lead to a temporary account pause. Accounts that break the violent-activity rules can still be banned outright.

Reporting and context

Alongside the changes, OpenAI introduced a Community Safety Report that it plans to update every six months, with the first edition due in early 2027 covering the second half of 2026, according to Unite.AI. The company says it made more than 200 referrals to law enforcement over potential threats of violence in the third quarter of 2026, and that it relies on mental health and behavioral experts for difficult cases.

The announcement follows a Common Sense Media report earlier in October that found ChatGPT's teen version often alerted parents late, or not at all, when teens discussed suicide, self-harm or eating disorders. OpenAI has not said how it will measure whether the new limits work.