OpenAI lets ChatGPT end violent chats, suspend repeat accounts
OpenAI says ChatGPT can now automatically end conversations focused on violence against others.
· Source: Reuters · Last verified: 31 Oct 2026
Summary
- OpenAI says ChatGPT can now automatically end conversations focused on violence against others.
- Accounts that trigger multiple such endings may be temporarily suspended, the company said.
- Teens lose access to practical weapons information even when questions appear innocent.
The latest
ChatGPT will be able to end a conversation on its own when a user shows a serious focus on violence or on harming another person, OpenAI said on October 30, 2026, and accounts that trigger multiple such endings may be temporarily suspended. The company said the safeguards apply to users of all ages and begin rolling out over the coming weeks.
Details
- The trigger: OpenAI said automatic termination applies when a conversation shows sustained focus on violence or on seriously harming another person. Repeated endings for that reason can lead the company to temporarily suspend the account. The company did not specify how many conversations constitute repetition.
- Teen restrictions: Users the system expects to be under 18 face added limits on weapons-related information even when a question appears innocent, according to the company. Practical details on weapons and on security systems that could ease violence are restricted, with or without any signal of violent intent.
- What teens keep: OpenAI said minors can still get help preventing or de-escalating violence, seek support, and be guided toward a trusted adult. Questions about the impact of violent events, warning signs and prevention remain available, the company said.
- Mass attacks: Responses about past mass attacks and their perpetrators will carry fewer details regardless of the user's age or stated reason, according to OpenAI. The withheld material includes weapons and equipment used, how they were obtained, and comparisons of the scale of harm caused.
- Age disputes: Users wrongly placed in the under-18 experience can confirm adult status through Persona, OpenAI's age-verification partner. The company said the strictest restrictions apply whenever the system predicts a user is a minor, defaulting toward the tighter setting.
- Parent alerts: Parents may be notified if a teen's account is disabled over violence-related violations, the company said. OpenAI added that it is building alerts for cases where an account is temporarily suspended because of repeated violent conversations.
- The numbers: OpenAI reported more than 200 referrals to law enforcement in the third quarter of 2026. It also committed to a community safety report covering violence prevention and the effectiveness of its safeguards, with the first edition due in early 2027 and releases every six months after.
- The timeline: The teen ChatGPT experience launched on August 18, 2026, according to the company. Age prediction began on consumer plans on January 20, 2026 and started in the European Union per an August 25, 2026 update, while parental alerts were expanded in July 2026.
- Outside findings: Reuters reported that OpenAI's own first study on teen usage, published earlier in October, found teens average under 15 minutes a day on ChatGPT. A Common Sense Media report that month found the teen version often alerted parents late, or not at all, when accounts discussed suicide, self-harm or eating disorders.
- Also: Change extends enforcement from teen safeguards to users of every age.
Between the lines
The enforcement logic has shifted. Earlier measures restricted what ChatGPT would say; automatic termination and account suspension act on the user instead. The Common Sense Media finding on missed parental alerts and the 200-plus law enforcement referrals both point to the same gap this package targets: detection that arrives after the risk, not during it.
What's next
Watch the rollout over the coming weeks, and the first community safety report due in early 2027 covering the second half of 2026, which OpenAI said will include anonymized examples, new policy changes and supporting data.