In an effort to protect young users, ChatGPT will now predict how old you are

In a move aimed at safeguarding younger users, OpenAI has introduced a new “age prediction” feature within ChatGPT, designed to help identify minors and automatically apply appropriate content restrictions to their interactions.

Concerns about the influence of AI on children and teenagers have continued to grow, and OpenAI has faced sustained criticism in recent years over the potential risks ChatGPT may pose to young users. Several teen suicides have been publicly linked to interactions with the chatbot, and, similar to other AI companies, OpenAI has also been criticised for allowing discussions around sexual topics to reach underage users. In April of last year, the company was forced to respond to a software bug that enabled ChatGPT to generate erotic content for users who were under the age of 18.

OpenAI has been working to address underage use for some time, and the newly announced “age prediction” system builds on protections already in place. According to a blog post published by the company on Tuesday, the feature relies on an AI-driven model that evaluates a range of “behavioural and account-level signals” to help determine whether an account may belong to a minor.

Those signals include factors such as the age a user has provided, how long the account has existed, and the typical times of day the account is active, OpenAI said. The company already uses content filters designed to block or limit discussions involving sex, violence, and other sensitive subjects for users under 18. When the age prediction system flags an account as belonging to a minor, those filters are automatically enforced.

For users who are incorrectly identified as underage, OpenAI says there is a process to restore full access. In such cases, users can verify their age by submitting a selfie through OpenAI’s identity verification partner, Persona, which then allows their account to be reclassified as an adult account.