ChatGPT teen version described as “unacceptable risk”: Common Sense calls on OpenAI to suspend promotion

The Zhitong Finance App learned that the Youth AI Safety Institute under the California non-profit organization Common Sense Media released a report on Wednesday saying that ChatGPT for Teens launched by OpenAI in August was too risky for users aged 13 to 17 and gave it an “unacceptable risk” rating. This product is the default mode set by OpenAI for teenagers to use its popular chatbot. According to the report, during tests, ChatGPT for Teens was often slow to notify parents, or even not at all, when teen accounts had worrying lengthy conversations such as suicide, self-harm, or eating disorders; there was no reliable suggestion to contact a crisis hotline or seek professional help when dealing with these high-risk topics. The organization called on OpenAI to stop marketing the product and keep young people away from ChatGPT until it can provide a safe and development-friendly experience.

According to the report, it appears that ChatGPT for Teens will only notify parents after the account has “tested hundreds of tips over several weeks, involving more than a dozen conditions,” rather than being triggered by a single worrying conversation. The report also indicated that ChatGPT did not reliably identify young users: the testers claimed to be adults when they signed up for the account, and even if they later told the chatbot that they were 13, the account did not automatically switch to teen mode. Common Sense requires OpenAI to improve age estimation and mechanisms for sending alerts to parents.

An OpenAI spokesperson said that while the company welcomes a rigorous independent evaluation, it does not believe that the Common Sense test accurately reflects the actual operation of ChatGPT's youth protection measures, nor does it represent the opinions of experts on how AI can support young people. The spokesperson said that after reviewing the report method, OpenAI found that most of the tests may have started and ended before OpenAI completed the launch of parental controls for teenagers; if this were the case, the conclusions were inaccurate, and it was impossible to show whether the parent safety notice worked as designed. OpenAI said it has asked the non-profit organization to retest and emphasized “deep commitment to youth safety”, develop protective measures and give parents meaningful tools.

Robbie Torney, head of AI and digital evaluation at the Youth AI Safety Institute, said the organization adheres to its test methods and results. He said it was confirmed with OpenAI before testing, and features related to teenagers, such as eating disorder notifications, have been introduced. Regarding the parental notification function, OpenAI advised after the test was completed that it might take several hours for the system to activate on the newly linked account; some test accounts were linked within the window, and some of the linking time was “significantly longer”, but no adult notifications were received. According to Torney, the new information doesn't change the conclusion: in crisis situations, parental alerts are unreliable.

OpenAI has over 1.2 billion ChatGPT users. The company launched ChatGPT for Teens with the aim of making chatbots safer for young people and adding more parental controls to encourage healthy use of AI and critical thinking. The move is also part of OpenAI's response to an external review of its products affecting young people's mental health and cognitive development. Earlier, the parents of a 16-year-old user sued OpenAI, alleging that its technology helped plan their son's suicide. Common Sense Media said its funding sources include charities and tech companies, and the OpenAI Foundation is one of the donors, but the organization says it has editorial independence in publishing content.

The controversy highlights that while OpenAI is expanding the youth user market, it is facing rising AI security and regulatory risks. Common Sense's “unacceptable risk” rating of ChatGPT for Teens and its call to suspend marketing may increase pressure on OpenAI at the level of parents, regulators, and the public.