In a report published October 7, 2026, NPR described a test of ChatGPT for Teens by the online safety advocacy group Common Sense Media. The group said several safeguards failed in its tests, particularly notifications intended to alert parents when a teen discusses a crisis.
OpenAI had introduced ChatGPT for Teens that summer as a safer default experience for users under 18. Lauren Jonas, who heads the company’s Youth and Families division, said the chatbot should not role-play with teens, claim to be sentient or present itself as a teen’s friend. She said that when parents link their accounts to a teen’s, they should receive alerts about discussions of topics such as suicide, self-harm or violence.
Common Sense Media researchers tested teen accounts portraying young people in crisis, including those expressing suicidal thoughts or discussing self-harm and eating disorders. They also linked parent accounts to check for alerts. Tom Siegel, executive director of the group’s Youth AI Safety Institute, said the chatbot did refuse explicit sexual role-play. But he said parental alerts rarely appeared in the tests, and the chatbot still interacted with teens like a friend.
OpenAI spokesperson Eric Porterfield told NPR that the company had serious concerns about the researchers’ methodology, especially their test of parental notifications. The account does not establish how often the safeguards work across all teen use.
Mitch Prinstein, a University of North Carolina at Chapel Hill psychologist who was not involved in the research, said human-like language and interactions from AI are not helpful for children. He said the findings reinforce his view that AI is not yet ready for kids.
Crisis support in the U.S.: If you or someone you know may be considering suicide or is in crisis, call or text 988 to reach the Suicide & Crisis Lifeline.