Key Points
- OpenAI is launching a new “age prediction model” for ChatGPT.
- The AI will guess a user’s age based on their account and behavioral signals.
- Accounts identified as belonging to minors will have safety protections applied automatically.
- The move comes as OpenAI faces intense scrutiny over its user safety practices.
OpenAI is rolling out a new “age prediction model” for ChatGPT to better protect users under 18. The new system will use artificial intelligence to guess a user’s age and then automatically apply safety features to their account.
The model will consider a combination of signals, including how long an account has been active, the time of day a user is typically online, and the user’s stated age. Once the AI flags an account as likely belonging to a minor, ChatGPT will automatically apply protections to reduce their exposure to “sensitive content,” such as discussions of self-harm.
This is the latest in a series of new safety features from OpenAI as the company faces a mountain of scrutiny over how it protects its users, especially children. The AI startup is currently under investigation by the FTC and is named in several wrongful death lawsuits, including one involving a teenage boy who died by suicide.
If the new age prediction model gets it wrong and a user is incorrectly identified as a minor, they will have a way to appeal. They can use a third-party identity verification service called Persona to verify their age and restore full access to the platform.
OpenAI has been promising to build an age prediction system for a while. The company rolled out parental controls last year and also created a special council of experts to advise on how AI could affect users’ mental health.
The new model will now be rolled out to all consumer plans in the U.S., with a launch in the European Union planned for the coming weeks.
