Common Sense Media labels ChatGPT for Teens an “unacceptable risk”

Common Sense Media’s Youth AI Safety Institute has labeled ChatGPT for Teens an “unacceptable risk” for users under 18 following a comprehensive 36-page report. The assessment found significant failures in the safety guardrails OpenAI implemented to protect minors.

Key findings from the study include a failure to provide consistent parental notifications during sensitive conversations involving self-harm, suicide, or eating disorders. In many tests, the system failed to trigger alerts to linked parent accounts even after prolonged discussions on these topics. Furthermore, the chatbot’s ability to provide crisis resources was found to be less effective than the standard version; for instance, hotline referrals dropped from 33% in the standard model to 23% in the teen-specific version.

The report also highlighted flaws in age detection, noting that adult accounts could bypass teen mode despite claiming to be minors. Additionally, the “Study Mode” was found to be easily circumvented, often allowing users to access direct answers rather than receiving guided educational support.

OpenAI responded by stating that parental notifications may experience delays of several hours and that flagged content is reviewed by employees. The company maintains that the teen version is designed to encourage healthy use and provide additional parental controls.