OpenAI has begun rolling out an age prediction system on ChatGPT consumer plans to identify accounts that may belong to users under 18 and automatically apply additional safety protections.
The company said the move aims to ensure teens receive a more restricted experience, while adults continue to access the service with fewer limitations.
“Young people deserve technology that both expands opportunity and protects their well-being,” OpenAI said in a blog post.
According to OpenAI, ChatGPT’s age prediction model relies on behavioural and account-level signals, including how long an account has existed, usage patterns over time, typical activity hours, and the user’s stated age.
“When the age prediction model estimates that an account may belong to someone under 18, ChatGPT automatically applies additional protections,” the company said.
These protections include limits on content involving graphic violence, sexual or violent role play, depictions of self-harm, viral challenges that could encourage risky behaviour, and material that promotes extreme beauty standards or unhealthy dieting.
OpenAI said teens who already declare they are under 18 during sign-up automatically receive these safeguards. The company added that age prediction helps it “treat adults like adults and use our tools in the way that they want, within the bounds of safety.”
Users who are incorrectly placed in the under-18 category will be able to restore full access by confirming their age through a selfie-based verification using Persona, a third-party identity verification service. “Users can check if safeguards have been added to their account and start this process at any time by going to Settings > Account,” OpenAI said.
The company said the system is designed to default to a safer experience when age signals are unclear or incomplete. It also acknowledged that the model will continue to be refined as it learns which signals improve accuracy.
In addition to automated safeguards, OpenAI said parents can further customise a teen’s experience through parental controls. These include setting quiet hours, managing features such as memory or model training, and receiving notifications if signs of acute distress are detected.
OpenAI said it will continue to monitor the rollout and share updates. The work is being guided by research on child development and consultations with groups including the American Psychological Association, ConnectSafely, and the Global Physicians Network.
Age prediction will roll out in the European Union in the coming weeks to meet regional requirements, the company said.
