A new evaluation questions parental alerts and other protections, while OpenAI challenges the testing methodology.
Common Sense Media says ChatGPT’s protections for teenage users fall short of the company’s promises, giving the service its “Unacceptable Risk” rating in an October 7 report.
The organization said some safeguards worked, including restrictions on explicit sexual role-play. But it identified problems with alerts intended to notify parents about conversations involving suicide, self-harm or eating disorders.
OpenAI disputed the findings. The company said much of the testing may have occurred before parental controls were fully activated and argued that the evaluation did not accurately represent its safeguards.
The disagreement comes as OpenAI promotes new tools intended to help teenagers study and plan for college.
In a separate announcement, the company said teenage users spend less than 15 minutes a day on ChatGPT on average. That usage figure does not establish whether individual conversations are safe.
Common Sense Media urged OpenAI to restrict access to adults until it can provide what the organization considers an appropriate experience for minors. OpenAI said it remains committed to improving protections and providing tools for parents.
Illustrative photograph; the person pictured is not associated with the evaluation. Photo by Viralyft on Unsplash.
