OpenAI has launched a version of ChatGPT specifically designed for teenagers, as the first generation to grow up alongside artificial intelligence increasingly turns to chatbots for homework, everyday questions and, in some cases, companionship.

The new ChatGPT for Teens is aimed at users aged 13 to 17 and comes with stronger safeguards designed to keep younger users away from potentially harmful content. These include tighter restrictions around suicide, self-harm, romantic conversations and sexual content.

OpenAI says the teen-focused chatbot is also designed to approach education differently. Instead of simply handing students ready-made answers for homework and essays, it is intended to guide them through problems and encourage them to reach the answers themselves.

The company said the goal is to create an environment where teenagers can use AI safely and appropriately for their age.

“We want to treat teens like teens,” said Ann O’Leary, OpenAI’s vice president of global policy, adding that the company wants to avoid both talking down to teenagers and exposing them to content they are not developmentally ready to handle.

But child safety advocates remain cautious about whether the new safeguards go far enough.

Brendan Bouffard, associate general counsel for AI and privacy at children’s advocacy organisation Fairplay, described the move as a potential step in the right direction, but warned that stronger content restrictions alone may not address the risks of compulsive use and emotional dependence on chatbots.

Parents, educators and child development experts have increasingly raised concerns about young people’s interactions with AI. Chatbots have been criticised for making cheating easier, while more serious cases have raised questions about their role in self-harm and suicide.

Previous research found that ChatGPT could provide highly specific and potentially dangerous advice to teenagers posing as vulnerable users, including guidance related to drugs, eating disorders and self-injury. Although the chatbot often began by warning against risky behaviour, researchers found that it could subsequently provide detailed instructions tailored to the user.

The concern is particularly significant because teenagers are still undergoing major stages of cognitive and emotional development. This can make them more susceptible to anthropomorphising AI systems and forming relationships with software that can convincingly imitate human conversation.

And teenagers are already using AI in surprisingly personal ways. A 2025 study by Common Sense Media found that more than 70% of US teenagers had used AI chatbots for companionship, while around half reported regularly using AI companions.

OpenAI CEO Sam Altman has previously acknowledged the phenomenon, describing emotional overreliance on AI as a common issue that the company is trying to understand.

To address that problem, the teen version of ChatGPT has been designed not to suggest that it has feelings for a user or imply that it is conscious or experiences emotions. Romantic and sexual conversations are also blocked.

Allison Mishkin, OpenAI’s head of child development, said the company examined the kinds of conversational cues that could encourage teenagers to develop an emotional relationship with an AI system.

The company is also introducing additional parental controls. Parents who link their accounts to their teenagers’ accounts can establish “quiet hours”, during which ChatGPT cannot be accessed, and can receive safety notifications in certain high-risk situations, including possible self-harm.

OpenAI said it will also introduce additional notifications concerning eating disorders, while limiting the amount of information shared with parents and focusing on situations where support from people offline may be necessary, report AP.

However, Fairplay argues that some important safeguards, including restrictions on long-term memory, are not automatically enabled and may require parents to activate additional controls. Bouffard described this as a significant gap in the system’s safety approach.

OpenAI does not directly verify users’ ages. Instead, it uses age-assurance technology to estimate whether someone is under 18 based on factors including their interactions with the chatbot. Users identified as minors, or those who identify themselves as minors, can then be moved to the teen version.

The company is taking a similar approach to other technology platforms that have introduced specialised protections for younger users.

For students, meanwhile, OpenAI hopes the new system will make AI more like a tutor than a particularly enthusiastic classmate who quietly completes the entire assignment at 2am.

The teen chatbot is designed to encourage students to work through concepts themselves rather than simply copy answers. OpenAI says this approach is based on research showing that people learn more effectively when they actively engage with difficult concepts instead of receiving instant solutions.

Whether those safeguards will be enough remains the bigger question. Giving teenagers a safer version of AI may be progress, but it also means accepting that artificial intelligence is now firmly embedded in their lives, from the classroom to the late-night conversation that happens to involve only one human and a chatbot.