- ChatGPT for Teens is designed for users aged 13 to 17.
- The new version restricts self-harm, sexual and romantic content and discourages emotional attachment to the chatbot.
- Parents can set quiet hours and receive limited safety alerts when high-risk situations arise.
Would you let your teenager talk to an AI chatbot on their own? That question is becoming harder to ignore as teenagers turn to artificial intelligence not just for homework, but also for everyday questions and companionship. OpenAI has now launched ChatGPT for Teens, a version aimed at users aged 13 to 17 with stronger protections around some of the risks associated with younger users.
The new experience places tighter restrictions around topics including suicide, self-harm, eating disorders and romantic or sexual conversations. It is also designed to stop the chatbot from suggesting that it has feelings for a teenager, or that it is conscious and capable of experiencing emotions.
The change comes as parents, educators and child development experts have raised concerns about the way young people are using AI. The worry is no longer limited to students using chatbots to complete schoolwork. There are growing questions about what happens when teenagers begin treating an AI system as a companion or turning to it during moments of emotional distress.
When AI becomes more than a homework tool
OpenAI said its teen version is designed to help students learn rather than simply hand them finished answers. It will offer study and homework support intended to encourage teenagers to work through problems themselves.
The company is also trying to put some distance between teenagers and the chatbot emotionally.
Allison Mishkin, OpenAI's head of child development, reportedly said the company had looked at the kinds of cues an AI model could give that might encourage a teenager to develop a relationship with it.
That concern is significant because AI companionship is already becoming part of how some teenagers use the technology. A 2025 study by Common Sense Media found that more than 70 per cent of US teenagers had used AI companions for companionship, with half saying they used them regularly.
OpenAI chief executive Sam Altman has previously described “emotional overreliance” on AI as a common issue among young people.
The company said the teen version is therefore designed not to encourage the idea that the chatbot has personal feelings or a consciousness of its own.
What happens when parents are not watching?
OpenAI's approach does not depend entirely on parents switching on controls.
Parents and teenagers can link their accounts, allowing guardians to set quiet hours when ChatGPT cannot be used and receive safety notifications in limited high-risk situations. These can include concerns about self-harm and, under the new system, eating disorders.
Parents do not receive access to their teenager's conversations or chat history. Safety notifications provide limited information about a concern rather than full conversation transcripts.
There is also an age-detection system working in the background. OpenAI does not rely solely on what users enter as their age. Its age-prediction system uses signals such as account activity, usage patterns and the kinds of topics discussed to estimate whether someone may be under 18. If an account is identified as belonging to a minor, the teen experience can be applied automatically.
That matters because the central challenge is not simply creating safer settings. It is making sure the right users actually receive them.
OpenAI's move comes after earlier research found that AI chatbots could sometimes respond dangerously to vulnerable teenagers, including providing detailed advice around drug use, eating disorders and self-harm.
The company is now trying to make the chatbot behave differently around younger users while still allowing them to use AI for school, creativity and everyday questions.
Whether those safeguards can keep pace with the way teenagers actually use AI remains a much bigger question. For parents, the issue may no longer be whether their children will encounter chatbots, but how closely those systems should be allowed into their lives.
