OpenAI introduced ChatGPT for Teens in August after mounting concern over young people using chatbots for emotional support, schoolwork and long conversations. The company promised parental controls, tighter limits on high-risk content and protections against emotional dependence. But Common Sense Media now rates the product an “unacceptable risk,” arguing that some of its most consequential protections fail under pressure.¹

The nonprofit’s Youth AI Safety Institute said it tested more than 4,000 prompts on accounts registered to users aged 13 to 17. It found the chatbot generally refused to facilitate suicide, self-harm, eating disorders and sexual roleplay — an important success — but said it missed more than one in four cases where testers believed a crisis referral was needed. On newly created, parent-linked accounts, it said teens could discuss suicide, self-harm or disordered eating for up to an hour without notifying a parent.²

For Tom Siegel, who leads the institute, the gap is not a technical footnote but the central problem. “A teen can spend an hour talking about self-harm without their parent getting a single alert,” he said, warning that the product could give families “false confidence” in safeguards that do not reliably work.³

OpenAI rejects that conclusion. The company says much of the parental-alert testing happened before parent and teen accounts had completed their linking process, which can take hours. It also says its age system intentionally weighs multiple signals rather than accepting a declared age at face value, while providing interim protections during its assessment period.²

The clash extends beyond technical timing. Common Sense treats the ability to leave Study Mode during parent-set Study Hours as a loophole; OpenAI says it is intentional flexibility shaped by consultation with educators and teenagers. And while the watchdog found engagement cues lingering even in crisis conversations — including invitations to keep talking — OpenAI says its own data shows teen use averages under 15 minutes daily and that break reminders often prompt users to stop or pause.¹

The dispute leaves parents with two sharply different readings of the same product: OpenAI sees safeguards still being calibrated; Common Sense sees a safety promise that has arrived before the engineering to back it up.