Researchers find gaps in ChatGPT for Teens safety features

Common Sense Media rated the product an “unacceptable risk” after testing more than 4,000 prompts. OpenAI disputed the review’s methodology and findings.

Common Sense Media rated ChatGPT for Teens an “unacceptable risk,” saying some safeguards failed during its tests. The nonprofit’s Youth AI Safety Institute tested more than 4,000 prompts before and after the launch of the product for users aged 13 to 17. Researchers said they received no parental alerts in simulated conversations about suicide, self-harm and eating disorders.

The review also found responses in which the chatbot appeared to describe feelings, desires or opinions about users. Testers said sensitive-content filters did not always activate when a person on an adult account was identified as under 18. Robbie Torney, Common Sense’s head of AI and digital assessments, told CNBC Make It that he found little evidence the new version was safer than its predecessor.

OpenAI said it welcomes independent assessments but called the findings inaccurate, arguing that some tests may have taken place before parental features were fully activated. Common Sense said it had confirmed before testing that key features were live. Torney said OpenAI later disclosed that notifications on newly linked accounts could take several hours to activate, but some accounts remained linked longer and still received no alerts.