OpenAI has introduced ChatGPT for Teens, a new version of its AI chatbot designed specifically for users aged 13 to 17. The update, which began rolling out last week, automatically applies safety features to accounts it identifies as belonging to minors, aiming to provide a safer and more supportive experience for young users. This marks a shift from the previous model, where parental controls were available but required voluntary linking between parent and child accounts.
The company stated that its system uses various signals to estimate a user’s age, including conversation topics, activity times, usage patterns, and account duration. When an account is determined to belong to someone under 18, ChatGPT then activates protections intended to prevent harmful content exposure, especially regarding sensitive issues such as self-harm, eating disorders, graphic violence, and sexual or romantic role-play. The teen version also includes educational tools like study reminders and prompts to take breaks, and it discourages the chatbot from fostering emotional dependence by avoiding personification or positioning itself as a replacement for human relationships.
This change comes against a backdrop of widespread ChatGPT use among adolescents. Recent surveys indicate that nearly 60% of U.S. teens use ChatGPT, with a significant portion engaging without parental knowledge. In a separate study, approximately one in five individuals aged 12 to 21 reported using AI chatbots for mental health advice, with most not disclosing these interactions to anyone. OpenAI argues that automatic safeguards are crucial in reaching teens who might otherwise avoid parental supervision or fail to activate safety settings.
However, experts caution that the effectiveness of such automatic protections hinges on the accuracy of OpenAI’s age detection system—a point on which the company has not released detailed performance metrics. Similar age verification methods on platforms like Roblox have shown shortcomings, with reports of incorrect classification due to reliance on imperfect AI-powered tools.
Moreover, there are questions about the actual safety outcomes of these protections. Independent testing prior to the launch revealed instances where AI chatbots missed warning signs related to self-harm and even suggested inappropriate advice. OpenAI has published some performance data related to sensitive content categories but has not shared comprehensive evaluation details, such as testing methodologies or datasets, limiting independent verification.
Industry parallels raise additional concerns. Meta’s introduction of Teen Accounts on Instagram included automatic protections, yet independent research found that many features did not fully function as advertised. While Meta maintains that the settings reduce exposure to harmful content, the broader picture of effectiveness remains unclear due to a lack of publicly available data.
Observers emphasize that while OpenAI’s move to default teen safeguards is a positive development, transparency and rigorous, independent evaluation will be vital to assess whether these measures achieve their intended goals. OpenAI has pledged to publish ongoing learnings about the system but has yet to provide a detailed framework or timeline for such disclosures. Key questions remain about how reliably the system identifies teens, whether responses are safer in practice, and if the teen experience influences user behavior, including seeking human support during distress.
As OpenAI rolls out the new teen experience, stakeholders call for the company to facilitate third-party audits and regulatory oversight to ensure the technology delivers meaningful protection for younger users. Without such scrutiny, the company’s assurances risk remaining untested promises rather than demonstrated results.