If you opened ChatGPT this week and found that some things that used to work are now blocked, the likely reason is that you have been classed as a teenager. And the classification does not come from the birth date you entered when you signed up.
On 18 August OpenAI launched a “ChatGPT for teens” version and began rolling it out worldwide. What is new about it is not the restricted content but the way the system decides who is a teenager.
How the system guesses your age
The company calls the mechanism “age prediction”, and says it works independently of the birth date registered on the account. The signals it relies on, in the company’s own words: the general topics you talk about, the hours of the day you use the service, how and when the account is used, and how long ago it was created.
Which means a new account used at night on study topics may read as a teenager’s account, whatever the age of the person behind it.
What changes on a classified account
The account stays active and ad-free, but specific categories are blocked from it: graphic violence and gore, viral challenges that encourage risky behaviour, sexual, romantic or violent roleplay, and content promoting extreme beauty standards or unhealthy behaviours.
If you are an adult and were classed wrongly
There is one route: Settings, then “Account”, then “Age verification”. The verification is not carried out by OpenAI itself but by an outside company called Persona, using a live face capture or an official identity document, or both.
And the detail that matters to anyone hesitating before uploading a photo of their ID: the company says Persona deletes uploaded material within seven days, and that OpenAI receives only the age verification result, not the documents themselves.
And if you are the parent of a teenager
Linking is done from Settings, then “Parental controls”, then adding a family member, then inviting them by phone number or email.
What a parent can set: reduced sensitive content, which is on by default; consent for conversations to be used in model improvement; saved memory; voice mode; image generation; study mode settings and its schedule; and “quiet hours”.
What they cannot: read the conversations. The company states plainly that parental controls do not give access to a teenager’s chats, their history, or their live activity. The safety notifications sent to a parent do not include conversation text.
What the company has not announced
OpenAI has published no date for this version’s availability in Kuwait or the Gulf. The announcement says “globally” and names one country with a specific date, Australia on 8 September. It has also published no accuracy figure for age prediction, so how often it is wrong is unknown. Any error rate circulating did not come from the company.
Sources: OpenAI’s “ChatGPT for teens” announcement of 18 August 2026, and the official help centre pages on age prediction and parental controls, updated in recent days.
Kuwait One, once a week.
The most useful stories we published this week, in your inbox.
No spam. Unsubscribe anytime.
