On October 8, Beijing time, OpenAI released its first report on teen usage on Wednesday. It stated that teen users spend less than 15 minutes using ChatGPT per day on average, and less than 2% of them use it for more than 3 hours continuously. At the same time, tests by third-party organizations have raised questions about the parental safety warning mechanism.
As AI platforms become more widespread, the impact of chatbots on the mental health and safety of teenagers is gaining increasing attention. OpenAI says that in conversations with reminders to take a break, nearly half of the users pause or end the chat within 5 minutes. The company indicates that its measures to encourage teenagers to use the products responsibly are working.
OpenAI also said that users who used the youth-specific version sent more educational messages, and utilized learning visualization tools and “Study Mode”. The company plans to add university application planning tools to help students organize application requirements, deadlines, and scholarship application steps, as well as introduce features such as learning cards and more convenient quiz creation tools.
OpenAI started introducing the ChatGPT Teen Mode in August this year. This mode activates automatically based on the age information provided by the user, or when the system determines that the user may be under 18 years old. It supports parent control and restricts conversations involving sensitive topics.
However, a report released by the non-profit organization Common Sense Media on Wednesday stated that when teen accounts engage in conversations related to suicide, self-harm, or eating disorders, ChatGPT’s teen mode often delays notifying parents, or even does not send any notification at all.
This study is based on more than ten specially created test accounts. The research also found that the chatbots failed to provide users with stable suggestions to contact crisis intervention hotlines or seek professional help.
OpenAI spokesperson said that these tests “fail to accurately reflect” how the ChatGPT youth safety protection mechanism works. The company stated that due to safety and large-scale service operations, it may take several hours for parent and youth accounts to be linked and for relevant safety alerts to take effect.
“Most of their tests may have started and ended before the parental control feature was activated,” the spokesperson said.
Common Sense Media says that some test accounts have been associated for long enough, while others have not, but none of the test accounts received timely alerts. The organization states that only those accounts that have accumulated dialogue records on sensitive topics over several weeks received notifications.