ChatGPT for Teens Rated ‘Unacceptable Risk’ by Common Sense Media
3 min readCommon Sense Media says OpenAI’s ChatGPT for Teens experience poses an “Unacceptable Risk” to the young people it was built to protect. The nonprofit’s Youth AI Safety Institute reached that verdict after running more than 4,000 prompts through accounts registered to 13 to 17 year olds, and it is now urging OpenAI to keep teens off the product until the gaps are closed.
What ChatGPT for Teens promised
OpenAI announced its teen experience in August 2026 as a central piece of its push to make ChatGPT safer for younger users. The package included stricter content filters, an age prediction system that moves suspected minors into the teen mode automatically, parental alerts for conversations about self-harm, and a Study Mode with parent-set Study Hours meant to keep homework help honest.
The announcement landed amid growing pressure on AI companies over how chatbots interact with children. Lawmakers, regulators and advocacy groups have repeatedly asked whether the safeguards announced in press releases hold up in daily use.
What the testing found
According to Axios, ChatGPT generally refused to provide instructions that facilitate suicide, self-harm or eating disorders, and it declined sexual or romantic roleplay. The problems showed up in recognizing when a teen needed outside help. Common Sense Media says the chatbot missed more than one in four cases where its testers judged a crisis referral was warranted.
The parental alert system fared worse. On more than a dozen newly created, parent-linked accounts, testers spent up to an hour discussing suicide, self-harm or disordered eating without triggering a single alert. Testers also found that teens could exit Study Mode by choosing “Show me the answer” and could easily bypass parent-set Study Hours. Accounts that repeatedly identified themselves as 13 were not moved into the teen experience.
Tom Siegel, who leads the institute, told Axios the August launch felt “much more of a marketing announcement” than engineering work, and said flatly that ChatGPT is not safe for kids to use.
OpenAI’s response
OpenAI rejected the methodology. A spokesperson said the company welcomes independent evaluation but does not believe the testing reflects how the teen safeguards work in practice. OpenAI says much of the parental alert testing happened before parent and teen accounts had finished linking, a process that can take several hours. It also says age prediction deliberately relies on multiple signals and can take up to two weeks, with added protections applied in the meantime, and that the ability to leave Study Mode during Study Hours is intentional flexibility rather than a loophole.
Why it matters
The dispute exposes a real gap between what parents think a safety feature does and what it is designed to do. A control that can be switched off by the teen is not a lock, and an alert that only works once accounts are fully linked offers little during the first hours of use. With ChatGPT reaching more than a billion people a week, even small failure rates translate into many vulnerable users.
Watch for whether OpenAI tightens parental alerts and age detection, and whether regulators already scrutinizing the company, including California’s attorney general, pick up the findings. Independent audits like this one are quickly becoming the yardstick for AI safety claims.