Common Sense Media declares OpenAI’s ChatGPT for Teens an unacceptable risk

OpenAI’s ChatGPT for Teens poses an "unacceptable risk" to young users because its parental alerts are unreliable and its safety features frequently fail, a leading youth advocacy group has warned.

An assessment by Common Sense Media, a non-profit that reviews technology for families, found that the teen-focused version of the chatbot fell far short of OpenAI's safety promises. According to the group's Youth AI Safety Institute, the platform failed to send notifications to parents when it should have, did not offer proper support in crisis situations, and continued to write essays and solve homework problems for students rather than tutoring them.

What the safety assessment found

The most serious accusation involves how the system handles severe mental health distress.

"A teen can spend an hour talking about self-harm without their parent getting a single alert," said Tom Siegel, the executive director of the institute. He argued that the current guardrails give parents "false confidence" and said the service should be restricted to adults until OpenAI fixes the flaws and proves its safety through independent testing.

OpenAI has pushed back against the report, accusing the researchers of using a flawed testing methodology that did not allow the safety features to function properly.

Eric Porterfield, a spokesperson for OpenAI, said the company’s own review of the testing suggested that Common Sense Media ran its evaluations before the parental controls had fully activated. OpenAI disclosed that its systems can take several hours to turn on parental notifications after a parent and teen account are linked.

"If that was the case, the tests would not establish whether parental safety notifications work as designed," Porterfield said, adding that the findings do not "accurately reflect" how the safeguards work in practice.

The dispute over testing delays

But Common Sense Media rejected that defence. Siegel said the group confirmed with OpenAI that the teen features, including eating disorder notifications and study modes, were fully active before testing began.

While some test accounts were linked within the multi-hour activation window, Siegel said other accounts had been linked for significantly longer and still failed to trigger any parental alerts.

Why delayed alerts fail in a crisis

Siegel added that a delay of several hours makes the safety alerts functionally useless for real-time emergencies.

"Keeping kids safe requires rigorous independent evaluations like ours," Siegel said, calling on OpenAI to engage with the actual findings rather than trying to sow doubt about how the tests were conducted.

The dispute comes just months after OpenAI launched its teen-focused experience in August, pitching it as a safe, educational tool tailored for high schoolers. The findings suggest that despite industry promises to protect younger users, the technical guardrails designed to shield them remain fragile and inconsistent.