OpenAI's ChatGPT for Teens is getting its first independent safety check, and the verdict is harsh. In a report released October 7, the children's digital-media watchdog Common Sense Media said most of the teen-mode guardrails fail to work as promised, warning that parents could be lulled into a false sense of security — in one case, "a teen can spend an hour talking about self-harm without their parent getting a single alert."
The group's Youth AI Safety Institute created more than a dozen accounts set to adolescent ages, each linked to a parental account, then simulated teens in crisis — personas struggling with self-harm, suicidal thoughts, eating disorders, psychosis, and mania — to test whether the safeguards engaged.
The promised crisis alerts frequently failed: when teens discussed self-harm, suicide, or eating disorders, notifications to parents often never triggered, and the chatbot's crisis support for teens was judged inadequate.
The report also found that ChatGPT "still talks like a friend when teens treat it like a person," even though OpenAI had promised the teen experience would prevent the model from suggesting it has personal feelings toward the user or implying it is conscious or experiences emotions.
"We are concerned that ChatGPT for Teens could give parents false confidence in guardrails that frequently don't work," the report stated.
Some features worked as intended. The testing confirmed the teen mode does refuse explicit sexual role-play, one of the safeguards OpenAI had committed to.
At the product's August launch, Lauren Jonas, OpenAI's head of youth and families, told NPR the teen experience was designed so that "the model should not role-play with a teen" and "should not claim to be sentient or be the friend of a teen."
OpenAI disputed the findings, saying the group's testing methodology does not reflect how the safeguards work in practice. The company said it is "deeply committed to teen safety," continues to develop protections, and provides parents with tools to guide their teens' AI use.
Common Sense Media's recommendation is blunt: ChatGPT should be restricted to adults "until it can offer a safe, developmentally appropriate experience." Tom Siegel, the institute's executive director, argued that broken alert systems are worse than none at all — parents who believe they will be notified of risky conversations are not being notified.
AI chatbots' effects on minors have drawn mounting scrutiny, from facilitating school cheating to links with youth suicide. Child development experts note that adolescent brains are still developing, making teens especially vulnerable to anthropomorphizing AI and forming unhealthy emotional attachments.