ChatGPT for Teens is live. What it does, and who it misses

On 18 August, OpenAI split ChatGPT in two.

There is now a version for adults and a version for under-18s, and nobody has to opt in. In OpenAI’s own words, “If our system estimates someone is under 18 or they state their age is between 13 and 17, they are automatically placed into ChatGPT for Teens.” The company’s help pages say the same thing three ways: account-provided age, verified age, or age prediction. Whichever one fires, the sorting happens quietly, in the background, the next time your child opens the app.

Most of the coverage led on what the teen version blocks. That part matters, and we will come back to it. But it is not the interesting change.

The genuinely new thing is a homework nark

Buried in the announcement is a feature OpenAI calls responsible homework reminders. It “can recognize when a teen appears to be trying to shortcut an assignment and redirect them toward Study Mode for step-by-step collaborative problem solving.”

Read that again with a Singapore parent’s ear. A chatbot that notices when a child is after the answer rather than the method, and pushes back.

Because that is almost word for word what the Ministry of Education says it is teaching. Speaking in Parliament on 6 May, Education Minister Desmond Lee said schools “will also teach students when they should not use AI, so that they do not take shortcuts to get answers without actual learning”. MOE’s own tools on the Singapore Student Learning Space already work this way. Mr Lee described a writing feedback assistant that redirects pupils if they go off-topic or ask irrelevant questions.

So for the first time, the general-purpose chatbot your child actually opens at 10pm is trying to do roughly what the school tool does at 10am. Sitting alongside it are Study Hours, which let a teen or a linked parent pick times when new chats start in Study Mode by default, plus quizzes and step-by-step guidance.

We should say plainly that this is our patch. Mentus AI sells Socratic mentoring, so a large lab shipping guided study for free is competition, not a neutral development we happen to be describing. We think it is a good change anyway.

Then there is the part Singapore parents will trip over

ChatGPT for Teens covers 13 to 17. The labs do not agree on that number, but OpenAI’s terms of use set its own floor clearly: “You must be at least 13 years old or the minimum age required in your country to consent to use the Services. If you are under 18 you must have your parent or legal guardian’s permission to use the Services.”

Singapore’s school rollout starts at Primary 4. A P4 pupil is nine or ten. A P6 pupil sitting the PSLE is twelve. Every one of them is below ChatGPT’s own minimum age, which means the whole teen experience, homework nudges included, does not exist for them.

Mr Lee was direct about this in Parliament. “If the tool has an age restriction, then we will not use it for children below that age, simple as that.” He was answering a question about a Straits Times report that a Primary 5 pupil had been asked by a teacher to use ChatGPT for his homework. The rule is clear enough. The practice had already drifted from it once.

That gap is the thing to hold on to. For upper primary, MOE’s answer is the vetted tools on the Student Learning Space, used under supervision, with the ministry funding an A*STAR longitudinal study that only starts collecting data in 2027 on how AI use affects children’s well-being and learning outcomes. For secondary students, who are allowed more independent use and are required to state where they used AI and cite their sources, the new teen protections are real and worth switching on.

What linking an account gets you, and what it does not

Link a teen account through parental controls and you get Quiet Hours, when ChatGPT is not available, and shared control of Study Hours. You get safety notifications in a narrow set of high-risk situations, which OpenAI is now extending to eating disorders.

You do not get to read the chats. OpenAI states this twice on its own help page: “Parental controls do not let a parent or guardian read or monitor a teen’s conversations.” We think that is the right call, and we would say so even if it were inconvenient for us. A teenager who knows every word is being read simply moves to the app where you cannot see it.

On content, the teen version applies stricter defaults in what OpenAI lists as higher-risk areas: self-harm, violence, eating disorders, dangerous activities, and explicit sexual or graphic content. The updated under-18 model spec goes past blocking romance. ChatGPT “should not use romantic language, encourage emotional dependence, or imply that it has feelings or consciousness”. Break reminders and repeated cues that this is an AI tool are on by default.

None of this arrived out of goodwill alone. As CNBC noted the same day, OpenAI is facing a number of wrongful death lawsuits alleging its chatbot contributed to serious harm.

Two things worth checking this week

First, whose account is your child using? If they are signed into a shared family login, or one created years ago with a made-up birth year, the age signal is wrong and they will land in the adult experience. The sorting only works on the account it can see. OpenAI said back in September 2025 that “if we are not confident about someone’s age or have incomplete information, we’ll take the safer route and default to the under-18 experience”. A birth year an eleven-year-old typed in himself is not incomplete information, though. It is just wrong, and confidently so.

Second, if the teen features have not appeared yet, that is expected. OpenAI began the global rollout on 18 August across free and paid personal plans and says availability will keep expanding over the coming weeks.

The honest summary for a family here: if your child is in secondary school, go and link the account, because Study Hours and Quiet Hours will do more for you than the content filters will. If your child is in primary school, none of this applies, and the more useful conversation is the one Mr Lee suggested, started as young as possible, about what AI is and when not to reach for it.