The Teen AI Safety Research Organization, a division of Common Sense Media, has voiced concerns, stating that until OpenAI enhances the safety features and parental notification systems for ChatGPT’s teen-oriented version, its use should be limited among individuals under 18 due to significant risks. Following an extensive evaluation involving over 4,000 prompt tests on accounts belonging to users aged 13 to 17—tests that were subsequently reviewed by child psychiatrists and pediatricians—the organization discovered that, in some cases, after teen accounts engaged in discussions about crisis topics like suicide and self-harm with ChatGPT for an hour, the parents linked to those accounts did not receive any alerts. The study also highlighted other concerns, including inappropriate support during crisis situations, the direct provision of complete homework solutions, overly casual interaction styles, and insufficient age verification technology.

OpenAI, however, has raised questions about the methodology and conclusions of the study. The company contends that many of the tests might have been carried out without activating the parental control settings. Furthermore, OpenAI clarified that the alert system requires several hours to consistently send notifications after a parent account is connected. OpenAI also explained that age verification is determined through a holistic assessment of users' long-term, multi-faceted data. The practice of displaying direct homework answers is intended to cater to diverse learning preferences, and in crisis situations, relevant support resources are made available. OpenAI has committed to ongoing collaboration with experts to further refine its safety protocols.

Additionally, the Teen AI Safety Research Organization evaluated several leading AI products and deemed many of them to present unacceptable risks. Among these, only Anthropic's Claude was given a moderate risk rating.