OpenAI is making ChatGPT more cautious around violence, and teenagers will feel it first. Under-18 accounts will be refused practical help on weapons and security systems, including questions that seem innocent, while every user gets narrower answers about earlier mass attacks.
The company published the changes on 30 October and says they begin reaching users within a few weeks. Alongside them it promised a Community Safety Report, its first, in early 2027.
The tightening comes after a year of pressure over how chatbots treat minors. OpenAI launched ChatGPT for Teens on 18 August: users aged 13 to 17, and anyone its age-prediction system estimates is a minor, are moved into a separate experience with quiet hours, more frequent break reminders and a ban on romantic or dependence-building replies. Linked parents are notified about violent threats, and now about possible eating disorders too.
Legal pressure followed. On 14 October, eight US states (Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana) filed separate lawsuits in their own courts, alleging that OpenAI misled the public about how safe ChatGPT is for children. These are allegations, not findings, and the weapons rules arrive two weeks after the filings.
Chats can now be cut off
A new tool lets ChatGPT end a conversation by itself when a user seems fixated on violence or on badly hurting someone. Several ended chats can lead to a temporary pause on the account. Teens keep access to help on calming a situation, finding support or reaching an adult they trust.
For all ages, ChatGPT will not walk through how a past attack was carried out, what the attacker used, or how those items were obtained, and it will skip rankings of which attacks did more damage. OpenAI’s announcement says this holds whatever the user’s age or reason. Users can still ask about what an attack did to victims and their communities, or what warning signs look like.
Who counts as a teen
ChatGPT guesses age from behavioural signals. Adults wrongly moved into the teen experience can prove their age through the verification service Persona. OpenAI says the parent alerts it added in July for deactivated teen accounts will be widened to cover temporary pauses.
In the July to September quarter, OpenAI says it passed over 200 cases to law enforcement. The report due next year will cover July to December 2026 and then repeat every six months, with anonymised examples and data. OpenAI has not said whether the rules differ by country, and nothing in the announcement mentions Pakistan.
