OpenAI has explained how ChatGPT predicts whether an account belongs to someone under 18, a system that adds extra safety protections for teen users.

The age prediction model studies account signals over time. These include the topics a user discusses, typical times of use, usage patterns and how long the account has existed. Users can see their age category under Settings, then Account, then Age. The company says the model keeps learning and is expected to improve as it sees more data.

If an account is flagged as likely teen, ChatGPT switches on its Teen experience. This limits sensitive topics such as violence, extreme beauty standards and unhealthy dieting, while keeping the chatbot useful for learning and creating.

The system is not perfect. OpenAI says it can wrongly flag adults as teens. Adults who are misclassified can verify their age with a live selfie or a government ID.

Context: the explanation follows an independent report that rated ChatGPT an unacceptable risk for minors after safety tests.