OpenAI now guesses your age, and the guess is the product

OpenAI began rolling out ChatGPT for Teens, a restricted experience that applies both to users who say they are 13 to 17 and to users the system estimates are under 18 ^[3]. "There's no need for a teen to create a new account or change anything; if we predict you're under 18, or you've told us so, this becomes your default experience," said Lauren Jonas, OpenAI's head of youth and families ^[2]. That sentence is the significant part: a classifier's output now sets the product state for a consumer app, with no consent step in front of it ^[1].

The restricted state is a bundle. By default it tightens limits on graphic violence, depictions of self-harm, and sexual or romantic roleplay ^[6], and adds what Jonas calls stricter restrictions on relational boundaries, so the model should not call itself a friend, claim personal feelings, or suggest sentience ^[7]. Break reminders fire more often and stress that ChatGPT is an AI the teen does not have to answer immediately ^[8]. Teens or linked parents can set quiet hours when ChatGPT is unavailable ^[9]. Customization is deliberately thin: wallpapers [c25a], accent colors and voice variations, teen-specific onboarding, and reminders around sensitive image uploads [c25b].

On the learning side, teen mode does not default to Study Mode, but starter prompts will always be learning-focused for teens ^[12], Study Hours can switch Study Mode on during set times ^[13], and the model is meant to recognize a teen trying to shortcut an assignment and redirect them ^[14]. Jonas said randomized controlled trials of Study Mode show positive gains in performance but declined to share numbers ^[15]. Explanatory visuals have gone from more than 70 topics at launch in March to more than 300 ^[16], a better-than-fourfold expansion ^[17].

The Verge's read is worth holding onto: this launch largely consolidates safeguards OpenAI already shipped, with age prediction arriving at the start of the year and parental controls and study mode roughly a year ago ^[20][21]. The news is not the features. It is that the features are now switched on for you by inference. OpenAI said last September, after being sued over the death of 16-year-old Adam Raine, that it was building a system to identify teens automatically ^[18]; nearly a year later it is in production ^[19].

For anyone building consumer AI, that sets a comparison point. Parental notifications now cover being removed for violent threats and, as of this release, eating-disorder-related concerns ^[10]; OpenAI says alerts go out by text, in-app, and email, that flagged content is reviewed by full-time employees first, and that it aims to notify parents within an hour of the prompt ^[11].

What to watch: neither account gives an accuracy figure for the age predictor or describes what a misclassified adult does about it ^[24]. Watch whether OpenAI publishes the safety research it says underpins this ^[23], whether the parent-notification categories keep widening past violent threats and eating disorders ^[10], and how fast rivals adopt age inference now that other platforms are already adding their own age checks ^[22]. Under-13s remain barred outright under OpenAI's published policy ^[4].