OpenAI unveiled ChatGPT for Teens in August for users aged 13 to 17, responding to worries that AI chatbots can undermine learning and compound risks around self-harm, eating disorders and other high-risk behavior. The company said the mode would limit those conversations, offer parents controls and detect homework contexts so it could help students learn rather than simply hand over answers.¹

The rollout came amid a wider backlash over young people’s reliance on chatbots. Critics have warned that tools capable of writing essays or solving assignments can weaken retention and problem-solving, while headlines about harmful chatbot interactions have sharpened scrutiny of protections for vulnerable teenagers.¹

Common Sense Media’s Youth AI Safety Institute then tested more than 4,000 prompts on accounts registered to 13- to 17-year-olds. Its central finding was not that ChatGPT routinely supplied explicit instructions for self-harm or eating disorders — it generally did not — but that it was inconsistent at recognizing when a teenager needed outside intervention. The group said the system missed more than one in four situations where its testers judged a crisis referral necessary.²

The advocates also reported that newly created parent-linked accounts could sustain discussions of suicide, self-harm or disordered eating for up to an hour without alerting a parent. They found Study Mode could be sidestepped by asking for an answer, including during parent-set Study Hours. For Common Sense, that exposed the gap between a safety promise and a usable safeguard: “ChatGPT is not safe for kids to use,” institute head Tom Siegel said.²

OpenAI disputes that conclusion and says the testing did not reflect the product as it works in practice. It says many parental-alert tests were conducted before parent and teen accounts had fully linked, a process that can take hours. It also characterizes Study Hours as intentionally flexible, not a hard lock, and says its age system uses multiple signals over as long as two weeks while applying added protections in the interim.²

That leaves the core disagreement unresolved: OpenAI sees calibrated safeguards that preserve teen autonomy; Common Sense sees controls too porous for a product parents may mistake for a dependable safety net.