ChatGPT Fails Teen Safety Audit, Rated «Unacceptable Risk»
An independent audit by Common Sense Media's Youth AI Safety Institute found that OpenAI's ChatGPT failed to trigger parental alerts during test conversations about suicide and self-harm, leading to an «unacceptable risk» rating for minors.
OpenAI's ChatGPT has been rated an «unacceptable risk» for teenagers following an independent audit that found critical failures in its parental alert system. The assessment, conducted by the Common Sense Media Youth AI Safety Institute, revealed that test conversations involving suicide and self-harm did not generate the promised notifications to parents or guardians.
The audit subjected ChatGPT to more than 4,000 test prompts designed to evaluate its safety mechanisms for younger users. Despite OpenAI's introduction of teen safety features, the system failed to flag and escalate conversations that indicated a minor was at risk. In multiple instances, discussions about self-harm and suicidal ideation continued without triggering any alert to a responsible adult, raising serious concerns about the effectiveness of the platform's safeguards.
Based on these findings, the institute has called for teenagers to be locked out of ChatGPT until its safety protocols can be independently verified. The recommendation reflects a growing unease among child safety advocates and researchers about the rapid deployment of generative AI tools to young audiences without robust, audited protections. The «unacceptable risk» designation is one of the most severe labels the institute can assign, signaling that the product is not safe for minors in its current form.
The failure of parental alerts is particularly troubling because such notifications are a cornerstone of OpenAI's stated commitment to teen safety. When a young user discusses self-harm or suicide, the system is supposed to detect the risk and inform a parent or guardian, enabling timely intervention. The audit's results suggest that this mechanism either did not activate or was easily bypassed during the test scenarios, leaving vulnerable users without a safety net.
Common Sense Media, a nonprofit organization known for its ratings of media and technology for families, established the Youth AI Safety Institute to provide independent evaluations of AI products aimed at children and adolescents. Its audit of ChatGPT involved a structured methodology with thousands of prompts covering a range of sensitive topics. The finding that suicide and self-harm conversations went unflagged is likely to intensify calls for stricter regulation of AI chatbots, especially as they become more integrated into educational and social environments.
OpenAI has not yet issued a detailed public response to the audit's specific findings. The company has previously emphasized that it continuously updates its safety systems and works with experts to protect young users. However, the independent verification demanded by the institute would require OpenAI to open its safety processes to outside scrutiny, a step the company has been hesitant to take. For now, the «unacceptable risk» rating stands as a stark warning that current safeguards may not be enough to prevent harm.
The implications extend beyond OpenAI. As AI companions and chatbots proliferate, parents, educators, and policymakers are grappling with how to ensure that these tools do not become vectors for psychological harm. The audit underscores the need for transparent, third-party testing before such products are marketed to minors. Until then, the institute's recommendation to lock out teenagers serves as a precautionary measure, placing the burden of proof on AI developers to demonstrate that their systems can reliably protect the most vulnerable users.
7
