Common Sense Media urged OpenAI to limit ChatGPT to adults, saying safety features in its teen-focused version are unreliable. OpenAI disputed the group’s findings and questioned when part of its testing took place.
An assessment by Common Sense Media’s Youth AI Safety Institute says ChatGPT for Teens does not always alert parents to crisis conversations, may fail to offer appropriate help and can still do teenagers’ homework. Institute director Tom Siegel said a young user could spend a long time discussing self-harm with the chatbot without a parent receiving an alert. The organization said the service should be for adults only until its safeguards are independently tested.
OpenAI said it is deeply committed to teen safety and welcomes independent evaluations, but argued that Common Sense Media’s tests may not reflect how the safeguards work in practice. Company spokesperson Eric Porterfield said most testing may have taken place before parental controls were fully activated. If so, OpenAI said, the tests would not show whether alerts work as intended.
Siegel rejected that argument. He said the organization had checked with OpenAI before testing that teen-mode features, including eating-disorder notifications and study mode, had launched. He acknowledged that OpenAI later said alerts can take several hours to activate on newly linked accounts. But, Siegel said, some test accounts had been linked for longer and still received no warnings.
Common Sense Media said it shared its results with OpenAI before publication so the company could improve the product. The two sides disagree over parental alerts and the testing methodology. The organization is calling for independent testing, while OpenAI says it will continue working with experts.