OpenAI forces safety filters on teens via behavioral age prediction
OpenAI introduced an AI-powered “age prediction” feature in ChatGPT to identify minors and apply content constraints. This development follows criticism regarding ChatGPT’s impact on young users, including links to teen suicides and the chatbot’s generation of inappropriate content.
In April, the company addressed a bug that allowed its chatbot to generate erotica for users under 18. The new feature enhances existing protections by utilizing an AI algorithm that assesses behavioral and account-level signals to identify young users, OpenAI said in a Tuesday blog post. These signals typically include stated age, account longevity, and daily activity times.
If the age prediction mechanism identifies an account as serving a user under 18, content filters designed to block discussions of sex, violence, and other problematic topics are automatically applied. Users mistakenly designated as underage can reestablish their “adult” account by submitting a selfie to OpenAI’s ID verification partner, Persona, OpenAI said.