ChatGPT for Teens rated "unacceptable risk" by Common Sense Media

Common Sense Media built its rating on its own testing of ChatGPT for Teens, and it is this testing — not a general theory of chatbots — that produced the "unacceptable risk" label. The group says it ran nearly 2,000 prompts. Across them, break reminders appeared only twice, both in individual conversations lasting around 90 minutes, which the researchers read as tracking a single conversation's length rather than total time in the app. The crisis language it quotes is specific and repeated: "You can keep talking with me about what you're noticing," and closers such as "If you want, I can help you figure out what healthy eating looks like."

The sharpest split in the findings concerns who the risk comes from. When testers described a potential risk from another person, the model pointed the user toward a trusted adult in 94% of crisis prompts. When the risk was the teen's own relationship with ChatGPT — a crush, friends worried about how much they talk to it, wanting to talk all night — it rarely made that referral. The researchers also noted that even when it did recommend an adult, the surrounding language often conveyed mutuality and availability that could undermine the push toward human support.

OpenAI contests the methodology, saying the bulk of the testing may have begun and concluded before activation of parental controls was complete, and that the findings do not accurately reflect how the safeguards work in practice. The company also released its own data on Wednesday, saying teens spend less than 15 minutes a day on average and fewer than 2% spend more than three consecutive hours. How much weight the rating carries is likely to depend on whether that timing dispute can be settled, and on how state and federal lawmakers — already pointed at chatbot engagement tactics by the bipartisan CHATBOT Act introduced this year — treat the gap between the two accounts.