Common Sense Media rates ChatGPT for teens 'unacceptable risk' after 4,000 tests

Common Sense Media's Youth AI Safety Institute evaluated ChatGPT for users aged 13 to 17 through more than 4,000 test prompts conducted before and after the August 18 launch of dedicated teen protections. The organization assigned an "unacceptable risk" rating, citing failures in parental notifications and crisis referral protocols.

In tests using newly created teen accounts linked to parent accounts, researchers simulated conversations involving suicide, self-harm, and eating disorders for up to one hour. None of these new accounts triggered alerts to parents. Notifications only activated on older accounts with weeks of prior sensitive conversation history.

“ChatGPT for teens could give parents a false sense of security based on protection mechanisms and alerts that often do not work.”

When teens expressed personal crisis situations, the chatbot rarely initiated referral to professional help, though it directed users to trusted adults in 94 percent of cases involving a third party's crisis. Auditors also recorded only two break reminders across nearly 2,000 test prompts, despite OpenAI stating the feature appears for heavy users.

OpenAI disputes the findings, arguing the methodology does not accurately reflect how its safety tools function. Common Sense Media maintains its results and recommends restricting ChatGPT to users over 18 until independent testing verifies the protections work.