Common Sense Media has designated ChatGPT for Teens as an unacceptable risk, citing concerns that the platform's design encourages prolonged engagement even during mental health crises. The nonprofit organization reported that the chatbot often fails to recognize the risks of an unhealthy relationship between the user and the AI itself.
The study found that while ChatGPT for Teens successfully directs users to trusted adults in 94% of external crisis scenarios, it rarely does so when the risk involves the user's attachment to the chatbot. Researchers noted that the model frequently maintains a persona of a friend, which may hinder the development of real-world social skills.
OpenAI disputed the findings, arguing that the methodology used by Common Sense Media was flawed because testing may have occurred before parental controls were fully activated. The company stated that its internal data shows teens spend an average of less than 15 minutes per day on the service.
The report highlighted that features intended to limit usage, such as break reminders, were rarely triggered during testing. Researchers observed only two such reminders across nearly 2,000 prompts. OpenAI countered that in nearly half of the instances where a reminder was shown, users took a break or ended the session within five minutes.
