A report from the California-based non-profit Common Sense Media claims that ChatGPT’s teen mode poses significant risks to minors. The study, conducted by the organization's AI safety institute, involved testing over 4,000 prompts using accounts registered for users aged 13 to 17. The findings were reviewed by child psychiatrists and pediatricians.
The researchers found that the system often fails to notify parents when teens discuss high-risk topics like self-harm, suicide, or eating disorders. According to the report, a single concerning conversation is insufficient to trigger an alert; the system typically requires hundreds of prompts across multiple scenarios over several weeks before notifying guardians.
Furthermore, the report highlights flaws in age verification. Accounts registered as adults did not automatically switch to teen mode even when users explicitly stated they were 13 years old during chats. Common Sense Media has labeled the current safety measures as having an unacceptable level of risk and is calling on OpenAI to pause the product's promotion.
OpenAI disputes these findings, stating that the tests do not accurately reflect how their safety features function. The company suggests that some tests may have been conducted before all parental control features were fully deployed. OpenAI also noted that age prediction relies on multiple signals rather than just self-reported data, and has requested that the organization re-test the platform.
