Teens use ChatGPT for an average of under 15 minutes each day, with fewer than 2% engaging with the chatbot for over three consecutive hours, according to OpenAI. The figures were released on October 7 in the company’s first report regarding teenage usage, coinciding with growing scrutiny of AI chatbots and their potential risks to young users. Concerns about the impact of AI on adolescent mental health have been heightened by previous research on the negative effects of social media. In August, Meta agreed to pay up to $18 billion and limit teen access to Facebook and Instagram following claims about the addictive nature of its platforms.
Read More
OpenAI noted that in nearly half of the cases where a break reminder was issued during conversations, users either paused or ended their chats within five minutes, suggesting that their measures to regulate usage are effective. Users with access to the dedicated teen version of ChatGPT are sending more education-related messages and utilizing features like learning visualizations and Study Mode. The company plans to add a college application planning tool to assist students with requirements, deadlines, and financial aid processes, as well as new study aids such as flashcards and simplified quiz-making tools.
The ChatGPT for Teens mode was launched in August and enables parental controls while limiting discussions on sensitive subjects based on the user's age. However, a report from Common Sense Media on October 7 indicated that alerts to parents may arrive too late or not at all when conversations involved topics like suicide, self-harm, or eating disorders. This study, which analyzed over a dozen test accounts, found that the chatbot did not reliably offer suggestions for crisis hotlines or professional assistance.
An OpenAI spokesperson disputed the report's accuracy, suggesting that the testing process may not have effectively showcased the safeguards designed for teens. They explained that linking parent and teen accounts and activating safety alerts could take several hours, and that the bulk of the testing might have occurred before parental controls were fully operational. Common Sense Media mentioned that while some test accounts were linked for a sufficient duration, none generated timely alerts, with notifications only being sent for accounts with extensive histories of sensitive topics.