OpenAI announced on Tuesday that it will launch an age recognition model for its ChatGPT personal version to help the artificial intelligence company identify accounts of users under the age of 18. The company said that the operation of this model relies on the combined analysis of account data signals and user behavior data signals. Specific signals cover users’ long-term usage habits, account registration duration, daily active periods, and age information reported by users themselves.
As scrutiny pressure continues to rise in the area of user protection, especially minor protection, OpenAI has launched a number of new security features in recent months.
The artificial intelligence startup and other technology companies are under investigation by the Federal Trade Commission over concerns that its artificial intelligence chatbots may have a negative impact on children and teenagers. At the same time, OpenAI has been involved in a number of wrongful death-related lawsuits, one of which centered on the suicide of a teenager.
OpenAI pointed out that once the age recognition model determines that a user is under 18 years old, ChatGPT will automatically enable a protection mechanism to reduce the possibility of the user being exposed to "sensitive content" such as descriptions of self-mutilation.
If a user is mistakenly identified as a minor, they can use the identity verification service provider Persona to complete the verification to restore full access to the account.
In August 2025, OpenAI announced that it would launch a parental control function to help parents understand and guide how teenagers use ChatGPT. The following month, the company officially launched this parental control feature and revealed that it was advancing the development of an age recognition system.
In October of the same year, OpenAI formed an advisory board of eight experts to provide professional advice on the impact of artificial intelligence on users’ mental health, emotions, and behavioral motivations.
OpenAI said on Tuesday that it will continue to optimize the accuracy of its age recognition model.
In order to adapt to regional compliance requirements, the model will be officially launched in the EU region in the next few weeks.