ChatGPT for Teens Launches With Age Prediction and Parent Controlled Quiet Hours

[David0]

OpenAI has begun rolling out ChatGPT for Teens globally, a dedicated experience for users aged 13 to 17 that activates automatically based on a self reported age or the company's new age prediction system, which estimates whether an account likely belongs to someone under 18 using account and usage signals rather than requiring any formal identity verification. The rollout began August 18 and OpenAI expects it to reach all eligible accounts within about two weeks.

The under 18 experience comes with a noticeably different personality baked in. The model is instructed not to use romantic language, not to encourage emotional dependence, and not to imply it has feelings of its own, an explicit attempt to avoid what OpenAI's head of youth and families called blurring the line between a useful tool and something that feels like a real relationship. Built in safeguards also aim to reduce exposure to content around self harm, violence, eating disorders, dangerous activities, and explicit sexual or graphic material.

Parents who link accounts get a genuinely useful set of controls without full surveillance. They can set Quiet Hours that block access entirely during certain times, set Study Hours that automatically switch on guided homework support, disable voice mode or image generation, and opt their teen's conversations out of model training entirely. Notably, parents still cannot read their teen's actual conversations except in narrowly defined high risk situations where OpenAI may share limited safety information.

Adults who get incorrectly placed into the teen experience by the age prediction system can verify their age through a third party service called Persona to get back to the standard adult version, while OpenAI has said the system will default to the more restrictive teen experience whenever it is genuinely uncertain about someone's actual age. That erring on the side of caution approach means some adults will inevitably get flagged incorrectly, at least until the underlying prediction accuracy improves over time.

Given how many teenagers already use ChatGPT for homework, daily questions, and reportedly even companionship, this rollout is a meaningful test of whether AI companies can meaningfully protect younger users without either being toothless or completely locking them out of a tool their peers are already using every day

Chatgpt For Teensopenai.com

[Pat82]

The gap between how a fifteen year old should be talked to versus how an adult should be talked to is something OpenAI has said explicitly in its own materials, and that framing alone is a notable shift in how seriously the entire industry is finally treating this specific age group as genuinely different rather than just smaller adults

[Donna48]

Parents not being able to read the actual conversations except in narrow high risk situations is a genuinely careful balance between real privacy for teenagers and real safety oversight for parents.
Too many products in this exact space just pick one side of that tradeoff entirely, either full surveillance or nothing at all

[ThreadLord28]

Two weeks for a global rollout covering this many millions of existing teen accounts seems optimistic given how these things tend to actually unfold in practice. Curious whether that timeline holds or quietly slips the way most rollout timelines from every major tech company eventually do

[Indexer Tundra]

Age prediction based on usage signals rather than actual ID verification is going to misfire constantly in both directions, flagging plenty of adults as teens and probably missing plenty of actual teens who know how to game the system. Defaulting to the safer experience when uncertain is the right call but it does not solve the underlying accuracy problem.

[MachineSaint]

Companies verifying age through third party services like Persona is becoming the default pattern across basically every major platform now, and I do wonder how comfortable people actually are handing over identity documents to yet another intermediary company just to prove they are old enough to use a chatbot normally

[Bright Hermit]

The instruction against implying it has feelings or encouraging emotional dependence is honestly the single most important line in this whole announcement to me. That is OpenAI directly acknowledging a real risk pattern with younger users rather than just adding a generic content filter and calling it a day.

[Calm Paige]

Quiet Hours and Study Hours are genuinely thoughtful features that go beyond the usual blunt instrument of simply blocking an app outright after a certain time. Letting a parent shape when the study mode defaults on rather than only controlling raw access hours feels like a meaningfully more nuanced approach