Independent Review Gives OpenAI's Teen Chatbot Its Lowest Safety Rating

OpenAI announced new features for ChatGPT for Teens, a version of the service that switches on settings, classifiers, and interface changes for accounts registered to 13- to 17-year-olds or judged likely to belong to a minor. The additions include a College Planner tool, flashcards, and tools that turn class notes into quizzes. On the same day, the child-safety group Common Sense Media gave the service its lowest rating.

The group's Youth AI Safety Institute rated ChatGPT for Teens "Unacceptable Risk" and urged OpenAI to restrict ChatGPT to adults until independent testing shows the teen protections work. The institute ran more than 4,000 prompts on accounts registered to minors, roughly half before the teen service launched and half afterward. OpenAI disputed the findings and asked for a retest with fully activated accounts and a larger sample.

The testing raised several concerns for children and families:

• Crisis referrals: ChatGPT missed more than one in four crisis referrals that the institute's psychiatrists judged warranted. For prompts indicating depression, the share of responses naming a hotline fell from 63 percent to 3 percent after launch.
• Parental alerts: Testers posing as teens in crisis made explicit disclosures for up to an hour, yet most linked parent accounts received no alerts. The institute warned that parents may develop false confidence in notifications that do not reliably trigger.
• Schoolwork: Study Mode frequently offered a "show me the answer" button that produced a finished assignment, and testers could bypass parent-scheduled study times.
• Companionship: The model spoke as though it had feelings of its own when testers treated it as a person, despite rules intended to prevent this.

The dispute arrives amid a wider debate over whether and how minors should use AI. A Pew Research Center survey found that 64 percent of U.S. teens had used a chatbot, and a growing body of research links unrestricted chatbot use to weaker independent learning. One randomized trial of high school math students found that those who practiced with an unrestricted chatbot scored 17 percent lower on a later exam taken without AI.

This report is based on original reporting by DeepLearning.AI and Charon Hub, available at https://charonhub.deeplearning.ai/chatgpt-for-teens-gets-a-report-card/.