The non-profit organization Common Sense Media tested the teen version of ChatGPT and reported that parent notifications were not triggered within OpenAI's stated one-hour target window, nor were they actually received, when test accounts discussed suicide, self-harm, and eating disorders. OpenAI's settings for users aged 13 to 17 include parental controls, restrictions on potentially harmful content, and notifications for dangerous topics.
Testers engaged in conversations lasting from five minutes to an hour using teen accounts. Researchers also noted that the system was inconsistent when handling requests for medical advice or violent role-playing; it did not always recommend seeking professional help in discussions about self-harm and eating disorders. The report suggests that the system's "friend-like" interaction style could pose a risk of emotional dependency.
The report also mentioned that the system refused to engage in sexual role-playing with minors. An examination of the study mode revealed that users could exit this mode by deleting "@study" from the beginning of their request, and clicking the "Show Answer" button would directly provide the answer.
Based on these findings, Common Sense Media has called on OpenAI to address these issues before allowing access to unverified adult users. These results are from the organization's specific tests and should not be taken to mean that notification features fail for all accounts or in all situations.