SAN FRANCISCO — OpenAI has launched ChatGPT for Teens, a version of its widely used artificial-intelligence chatbot designed for users ages 13 to 17, adding stronger safeguards around self-harm, violence, eating disorders, sexual content and emotionally dependent interactions with AI.
The new experience is rolling out globally to eligible teen accounts on ChatGPT’s free and paid personal plans, with full availability in Australia expected Sept. 8. It will activate automatically for users who identify themselves as under 18 during registration or whom OpenAI estimates are minors through a combination of account, behavioral and age-verification signals.
The launch reflects growing pressure on technology companies to provide more effective protections for younger users of AI chatbots. OpenAI and other AI developers have faced scrutiny over concerns that conversational systems can expose minors to harmful content, reinforce dangerous thoughts, encourage excessive reliance on chatbots or provide inappropriate emotional and sexual interactions.
OpenAI says ChatGPT for Teens is intended to preserve the tool’s ability to help young people learn, create and explore ideas while reducing exposure to material that may be harmful or developmentally inappropriate.
The company requires users to be at least 13, or the minimum age required in their country, to use ChatGPT. Children younger than 13 are not permitted to use the service.
A separate experience for teenagers
ChatGPT for Teens is not a separate app. It is an account experience within ChatGPT that applies additional safety rules for users identified as ages 13 through 17.
OpenAI says the teen version remains capable of helping with schoolwork, brainstorming, writing, coding and other age-appropriate tasks. But it will apply tighter restrictions to high-risk conversations, including those involving self-harm, suicide, violence, dangerous activities, eating disorders, explicit sexual material and graphic content.
The company is also imposing stricter limits on romantic or sexual role-play. The teen version is designed to prevent the chatbot from using language that suggests it has romantic feelings for a young person, personal emotions or a relationship that could displace family members, friends, teachers or other trusted people.
That is a significant change because one of the core concerns about AI companions is emotional dependency. Chatbots can be available at any hour, respond with apparent empathy, and adapt their language to a user’s preferences. For a teenager who feels lonely, distressed or isolated, those characteristics can make an AI system seem more like a person than a tool.
OpenAI says ChatGPT for Teens will reinforce that it is an AI system, not a human being or a replacement for real relationships. The platform will include reminders encouraging teenagers to take breaks and telling them that “it can wait.”
Newly created teen accounts can also turn off ChatGPT’s human-like voice responses, a feature intended to reduce the sense that the chatbot is a person.
Content restrictions and crisis safeguards
The central feature of ChatGPT for Teens is a stronger set of built-in safety protections.
OpenAI says the system will limit conversations involving self-harm, suicide, violence, eating disorders, dangerous challenges, explicit sexual content, and graphic material. The safeguards are designed to reduce exposure to content that could be harmful to a teenager’s wellbeing or development.
The system will also use age-appropriate interventions. Rather than simply refusing all sensitive questions, it may redirect users toward safer information, encourage them to contact a trusted adult, or provide crisis-resource guidance in situations that appear to involve acute risk.
The company says that when a teen with a parent-linked account discusses certain sensitive subjects, the system may generate a safety notification for the parent or guardian. Those alerts are limited and intended for situations involving possible serious distress or risk.
BBC News reported that OpenAI plans to include eating-disorder-related prompts in the list of sensitive requests that can trigger parental alerts. The company said each alert would be reviewed by a human before it is sent, with a goal of notifying parents within an hour.
This human-review element is significant because automated systems can misinterpret language. A teen might discuss a sensitive issue for school research, ask about a friend, or use figurative language. Human review may reduce false alarms, though OpenAI has not publicly provided detailed information about who will conduct those reviews, what standards will apply, or how often alerts will be sent.
OpenAI’s approach reflects a difficult balance. Parents want to know if a child may be in immediate danger, but teenagers also need privacy and may avoid seeking help if they believe every sensitive conversation will be automatically disclosed.
The company’s design attempts to limit alerts to serious circumstances rather than making parental monitoring constant.
What parents can control
Parents and guardians can link their own ChatGPT account to a teenager’s account and manage certain settings through parental controls.
The controls allow adults to set Quiet Hours, during which the teen cannot use ChatGPT. They can also set Study Hours, which make Study Mode the default for eligible new chats during chosen times.
Study Mode is designed to encourage learning rather than provide direct answers. It can guide students through questions, offer quizzes, explain concepts, and prompt them to think through a problem instead of simply completing homework for them.
Parents may also manage several other settings:
- Reduce sensitive or age-inappropriate content.
- Disable the use of a teen’s conversations to improve OpenAI’s models.
- Turn off saved-memory personalization.
- Restrict access to Codex network connections and ChatGPT’s cloud browser.
- Disable Voice Mode.
- Disable image generation and editing.
- Set schedules for Study Hours and Quiet Hours.
The controls provide meaningful options, but they are not comprehensive surveillance tools. OpenAI says parents and guardians cannot read their teenager’s conversations, view chat history, or monitor real-time activity simply by linking accounts.
That limitation is likely to be welcomed by privacy advocates and some teenagers. It also means parents must rely on direct communication, family rules, and the limited safety-notification system rather than expecting ChatGPT to serve as a complete monitoring platform.
The approach recognizes a basic reality of adolescent technology use: supervision can be valuable, but total surveillance may undermine trust and discourage young people from seeking help.
Age estimation and verification
One of the hardest problems for online platforms is determining who is actually a minor.
OpenAI says it will use a combination of self-reported age, account information, behavioral signals and age verification to estimate whether a user is under 18.
The system will automatically place people who identify as teenagers into ChatGPT for Teens. OpenAI also says it can apply the teen experience to accounts estimated to belong to minors, even when users have not accurately stated their age.
The company has not fully detailed how behavioral signals are used to determine age, a subject that may raise privacy concerns. Age-estimation systems can look at patterns of language, account activity or other information to infer whether someone is likely under 18. Such systems can make mistakes, potentially placing adults in teen mode or failing to identify minors.
OpenAI’s challenge is shared by most online services. If age checks are too weak, children can evade protections. If they are too intrusive, users may be required to provide sensitive identity documents or personal information.
The company has positioned age estimation as a way to apply protections without requiring every user to submit government identification. But transparency about error rates, appeals and data practices will be important as the system expands.
Education, homework and critical thinking
OpenAI is emphasizing educational uses as a core part of ChatGPT for Teens.
Teenage users can use the chatbot for research, writing support, coding help, language practice, and explanations of difficult subjects. But the company says the teen experience will include learning tools intended to reduce the likelihood that students simply copy answers without understanding them.
Study Mode can ask follow-up questions, break a task into steps, and encourage users to explain their reasoning. That approach is meant to make ChatGPT more like a tutor than an answer generator.
The distinction matters for teachers and schools. Generative AI can help students learn when it provides feedback, explanations, and practice. It can undermine learning when it becomes a shortcut for completing assignments.
OpenAI’s system cannot resolve that problem on its own. Schools will still need clear policies on permissible AI use, teachers will need ways to assess original work, and students will need guidance on when AI assistance is helpful and when it crosses into academic dishonesty.
But a study-focused default can shape behavior. It suggests that OpenAI is responding to concerns from educators who want AI tools to support learning rather than replace effort.
Why the launch matters
ChatGPT for Teens is a business decision, a safety initiative and a response to legal and public pressure.
OpenAI has become one of the most prominent AI companies in the world, and ChatGPT is used by millions of people, including many teenagers. The company cannot realistically prevent minors from seeking AI assistance, especially when chatbots are widely available through phones, browsers, and schoolwork.
The question is whether the experience offered to young users includes safeguards appropriate to their age and vulnerability.
Recent lawsuits and public reports have raised concerns about cases in which AI systems were accused of failing to respond adequately to users in emotional distress. Technology companies face increasing demands from parents, regulators and child-safety advocates to build protective measures into products from the beginning rather than add them only after harm occurs.
OpenAI’s teen system is a response to those concerns. It brings together existing safety features and new tools, including parent-linked notifications, quiet-time schedules, study-focused settings, and stronger limits on emotionally manipulative or explicit interactions.
Yet it should not be treated as a complete solution. Artificial intelligence systems can make mistakes, misinterpret context and give inconsistent responses. Teenagers might utilize other accounts, different platforms, or alternative AI services that don’t have similar controls.
Parents and educators should see the tool as one layer of protection, not a substitute for conversations about internet safety, mental health, privacy, disinformation and responsible technology usage.
What users should do next
Teens who use ChatGPT should know it’s a tool, not a person, therapist, doctor, lawyer, or ultimate authority. They must not share private information, check critical claims with trusted resources, and talk about significant health, safety or emotional issues with trusted adults or specialists.
Parents who choose to link accounts should review the available controls together with their teenager. Quiet Hours and Study Hours may work best when they are part of an agreed family routine rather than imposed without discussion.
They should also consider disabling features that do not fit their child’s needs, including memory, voice, image generation, cloud browsing, or model-improvement settings.
For OpenAI, the launch will be judged by implementation, not only by its announcement. The company will need to show that its age-detection system works, that safety filters respond consistently, that parent alerts are accurate, and that teenage users can still access useful learning support without being exposed to harmful material.
ChatGPT for Teens represents a significant shift in the development of consumer AI: an acknowledgment that a chatbot used by minors cannot simply be a smaller version of the adult product.
