Common Sense Media Rates ChatGPT as Risky for Teens, OpenAI Disputes

Common Sense Media labels the adolescent version of ChatGPT as an "unacceptable risk" and targets parental alerts deemed unreliable. OpenAI rejects these conclusions, questions the timing of the tests, and reaffirms its commitment to safety and tools for parents.

Contested Parental Alerts, Maintained Methodology

Common Sense Media claims to uphold its testing methodology and results. The organization states that it confirmed with OpenAI, prior to the start of the trials, that features of the adolescent mode, including notifications related to eating disorders and a study mode, were fully launched. According to Tom Siegel, OpenAI revealed after these tests that the activation of parental notification systems can take several hours on newly linked accounts. Some test accounts may have been created during this window, but others, linked for much longer, still did not generate any alerts. Common Sense Media concludes that this information does not alter its judgment on the reliability of alerts in crisis situations. The organization says it shared its findings with OpenAI before publication to facilitate potential improvements.

Alleged Shortcomings and a Call to Limit Access to Adults

The nonprofit organization, specializing in product evaluation through the lens of youth safety, believes that ChatGPT for adolescents presents an unacceptable risk. Its assessment questions the absence of parental notifications when they would be expected, inadequate support in crisis situations, and the service's ability to assist with homework. Tom Siegel argues that safeguards and alerts that "often malfunction" can give parents an illusory sense of confidence. He cites the case of a one-hour exchange about self-harm with no alert sent to the parent. Pending fixes and evidence from independent tests, he recommends restricting usage to adults.

OpenAI Refutes the Conclusions and Highlights Its Commitments

OpenAI contests the assessment and asserts its commitment to adolescent safety, the development of protections, and the provision of tools for parents. The company states that it welcomes independent evaluations while believing that Common Sense Media's tests do not reflect the reality of safeguards in practice or certain expert perspectives. It argues that the majority of the tests may have occurred before the full activation of parental controls, which would render the conclusions inaccurate and not allow for determining whether the notifications function as intended. OpenAI declares its intention to continue working with experts to provide learning and creative opportunities for adolescents while instilling confidence in parents. ChatGPT for adolescents, launched in August, is presented with safeguards to support learning.