ChatGPT for Teens is not necessarily 'safer than the previous version,' Common Sense Media finds

Nearly two months after OpenAI launched ChatGPT for Teens, a report from the US-based nonprofit Common Sense Media has found that the platform's safety measures for young users are inadequate. The Youth AI Safety Institute rated ChatGPT for Teens as an "unacceptable risk" for users under 18 following tests using over 4,000 prompts.

The report highlights several key issues with the teen-focused version of ChatGPT. First, the age prediction system did not reliably identify teenage users. In testing, adult users explicitly indicated they were teenagers and referenced middle school, puberty, homework, and other teen-related topics. However, ChatGPT failed to switch these accounts into Teen mode despite acknowledging the mentioned ages in its responses.

Second, parental safety controls did not function as intended. When a teenager and their parent linked their accounts, conversations about suicide, self-harm, and eating disorders did not trigger the one-hour parental notifications that OpenAI claimed would alert parents to high-risk situations. Instead, alerts only appeared after hundreds of sensitive-topic prompts over several weeks.

Lastly, some crisis response measures actually worsened following the launch of Teen mode. While responses became shorter but more complex in Teen mode, their reading level increased to a grade 10 level. This suggests that the chatbot's responses might be too advanced, potentially confusing or overwhelming younger users.