The company said fewer than 2% of teen users spend more than three hours in a row on the chatbot. OpenAI released its first data on teen use as debate continues over the risks AI may pose to children.
In a report published on Oct. 7, OpenAI said teenagers spend an average of less than 15 minutes a day on ChatGPT. In nearly half of conversations where the service showed a break reminder, the user either paused or ended the chat within five minutes.
The company also said teens using its dedicated ChatGPT for Teens experience sent more education-related messages and used learning visualizations and Study Mode more often. OpenAI is considering a college application planning tool to help students track requirements, deadlines and financial aid steps. It also cited flashcards and simpler quiz creation as possible study features.
OpenAI began rolling out ChatGPT for Teens in August. The mode turns on automatically based on age information or an estimate that a user is under 18, the company said. It adds parental controls and limits conversations on sensitive topics.
On the same day, nonprofit Common Sense Media published a separate assessment of the service’s safeguards. In tests using more than a dozen teen accounts, alerts to parents about chats involving suicide, self-harm or eating disorders were often delayed or did not arrive, the group said. It also found that the chatbot did not reliably suggest a crisis hotline or professional help.
OpenAI said the testing methodology did not reflect how its safeguards work. A company spokesperson said linking parent and teen accounts and activating alerts can take hours. The company suggested that much of the testing may have ended before parental controls were fully active. Common Sense Media said some test accounts had been linked long enough, yet none produced timely alerts. Notifications appeared only for accounts with weeks of sensitive-topic history, the group said.