OpenAI introduced a safety package for teenage ChatGPT users on July 16, 2026, which combines automatic age verification, break reminders, and new parental controls. According to the company, nearly nine out of ten teenagers already use the chatbot weekly for learning. Parents will soon be able to restrict their children’s quiet hours, speech mode, and image generation.
Age verification automatically detects users under 18
A new detection system estimates whether a person may be underage based on usage behavior and then automatically unlocks stricter default settings. Affected accounts will thus receive stricter protections against graphic violence, self-harm, risky challenges, and content related to unhealthy body images. Certain role-playing features will also remain restricted for detected youth accounts.
Additionally, the application displays reminders during long sessions that suggest taking a break. OpenAI is responding to criticism regarding excessive usage duration among young people, as has previously affected other chat and social media services. The company acknowledges misclassification of age as a technical limitation but does not publicly disclose exact error rates.
If the system is unsure, it opts for the more cautious approach and treats the account as that of a minor user. Adults will be able to verify their age in such cases to regain access to unrestricted features. Since September 2025, stricter default settings have applied to accounts identified as underage.
Parents receive extensive control features
By linking their account with that of a child, parents will soon be able to set quiet hours, disable speech mode, and control access to image generation. The memory function can also be turned off, and parents can exclude their child’s conversations from training future models. In cases of self-harm indications, trained staff will assess the situation before informing parents via email, SMS, or push notification.
To link accounts, a parent sends an invitation to the child’s account, which the child must confirm; if the teenager later disconnects, the parent will be notified. Without an active link, only the existing automatic protection measures for identified youth accounts will remain effective.
These features complement earlier protective measures that OpenAI introduced in response to growing political pressure in the U.S. Parental associations and regulatory authorities have repeatedly called for chatbot providers to implement warning systems for crisis situations and not rely solely on self-reports regarding age. As reported by Digital Trends, the new features are part of a longer series of previous youth protection announcements from OpenAI.
Learning tools and partner organizations complement the package
The core of the educational offering is the learning mode, which explains tasks step by step instead of providing direct solutions; it was developed in collaboration with teachers and education professionals. Supplementary starter prompts help to present topics clearly, convert notes into learning summaries, and create flashcards. Interactive learning offerings in mathematics and science reportedly reach 18 million users weekly across more than 250 topics, and a pronunciation tool supports over 61 languages – independently unverified.
For development, OpenAI collaborated with the American Federation of Teachers, the American Psychological Association, Common Sense Media, and the Family Online Safety Institute. The violence prevention organization Moonshot also advises on risks from radicalization and online violence. Its CEO, Vidhya Ramalingam, states that parents need “the right information to support their children.”
The announcement is part of a broader family strategy. According to an analysis by TechCrunch, nearly a quarter of all parents with smartphones in the U.S. now use ChatGPT, while parents tend to underestimate their children’s actual usage.
It will be crucial whether the behavior-based age estimation works reliably in practice, as similar systems from other platforms have been circumvented in the past. It also remains unclear whether the notification function for self-harm will actually reach families in time or if teenagers will turn to less regulated alternatives. OpenAI has not yet provided specific figures on the accuracy of the age verification.
