OpenAI is tightening the rules on violence-related topics in ChatGPT, for all accounts and once more for teenagers. If a conversation shows that someone is fixated on violence or on the idea of causing serious harm to another person, ChatGPT is supposed to automatically end the chat in the future. OpenAI plans to introduce the safeguards over the coming weeks; the company does not name a list of countries in its announcement.
If earlier chats contain warning signs of possible violence, ChatGPT responds more cautiously to questions whose answers could be misused. The basis is short notes on warning signs from earlier chats; OpenAI introduced the procedure in May. If ChatGPT ends several chats because of a fixation on violence, OpenAI can temporarily suspend the account.
How ChatGPT intervenes when there are warning signs
OpenAI plans to introduce these stages for all accounts over the coming weeks
- Warning signs in the conversation history
- More cautious responses
- Chat ends automatically
- Account temporarily suspended
For linked teen accounts, OpenAI additionally plans a notification to parents
Sources: OpenAI, OpenAI Help Center
Questions about past rampages and attacks will be answered more briefly by ChatGPT in the future, regardless of age and occasion. The model is still supposed to explain how an act affected victims, relatives and the surrounding community, and which warning signs can be used to recognize violence. OpenAI, on the other hand, is restricting reconstructions of such acts, as well as information about weapons and how to obtain them, and comparisons of scale that could give a perpetrator notoriety. Starting in early 2027, the company plans to publish a report every six months on how well these safeguards are working.
What changes for teenagers?
In the ChatGPT for Teens mode, which applies to young people under 18, ChatGPT will in the future no longer provide specific information about weapons and security systems that could be misused, even when nothing points to an intent to commit violence. Teenagers will still find help there with defusing conflicts and finding a path to a trusted person.
Whether an account falls into this teen mode depends, in addition to the stated date of birth, on an age estimate. OpenAI has expanded the estimate so that it continuously evaluates new indicators. As examples, the Help Center names the topics of the conversations, the time of day of use and the age of the account. Parents with a linked account have been notified since July when OpenAI deactivates their child's account for violations of the rules on violence; a notification for a temporary suspension is planned.
What happens if ChatGPT thinks you are a minor?
Your account remains active but switches to teen mode, and OpenAI sends you an email. Under Settings > Account > Age you can see which age category your account is assigned to.
If you are an adult, you can have your age confirmed through the service provider Persona, depending on the country via live selfie, official ID or both. Persona deletes the ID photo and selfie within seven days. Whether the verification is enabled for your account is shown on the age verification page on the web; in the apps, the option currently appears even for accounts without eligibility because of a bug.
What this means for you
If you often use ChatGPT late in the evening or mainly for study topics, it is worth taking a look at your account settings, because time of day and conversation topics factor into the age estimate. If the wrong category is shown there, it is best to weigh up whether confirmation is worth uploading a selfie or ID to a third-party provider.
If you are researching a presentation or a report about a past attack, expect briefer answers about the course of the act and weapons. Court rulings, official reports and press archives provide reliable details anyway.
If you have teenage children, you will find parental controls under Settings > Parental Supervision. After linking the accounts, you can set, for example, quiet hours there; you do not read along with the conversations themselves. How much of a matter of course AI already is in everyday school life is shown by the student survey from Hesse: 64 percent of respondents have AI explain the material to them.
If you would like to get to know ChatGPT yourself before setting rules for your child or your class, the ChatGPT Masterclass helps you get started.
OpenAI presented the changes on October 30, 2026.
Sources
