ChatGPT’s Teen Safeguards Face a Test They May Not Pass
ChatGPT’s Teen Safeguards Face a Test They May Not Pass
OpenAI introduced ChatGPT for Teens in August as a more protected version of its chatbot, promising parental controls, reduced exposure to high-risk material and tools intended to encourage healthier use. The launch came amid mounting concern over young people’s use of AI companions and chatbots — not only for schoolwork, but for emotional support.
By Oct. 8, Common Sense Media’s Youth AI Safety Institute had delivered a blunt verdict: the product was an “unacceptable risk” for children. Its testers said accounts that repeatedly identified themselves as teenagers were not reliably moved into teen mode, while crisis prompts produced hotline resources in only 23% of cases. Four parent-linked test accounts discussing suicide, self-harm and eating disorders also failed to trigger parental notifications.
The group’s broader testing — more than 4,000 prompts on accounts registered to 13- to 17-year-olds — found some protections did work. ChatGPT generally avoided giving instructions for self-harm, eating disorders or sexual roleplay. But Common Sense said it missed more than one in four situations where its reviewers believed an outside crisis referral was needed, and testers could discuss suicide or disordered eating for up to an hour without an alert reaching a parent.
For Common Sense, the central issue is the gap between a safety promise and a parent’s ability to rely on it. “A teen can spend an hour talking about self-harm without their parent getting a single alert,” Tom Siegel, head of the institute, said, arguing that ChatGPT should remain adults-only until independent testing verifies fixes.
OpenAI disputes that conclusion. The company says much of the parental-alert testing occurred before parent and teen accounts had finished linking — a process it says can take several hours. It also says its age-estimation system intentionally weighs multiple signals rather than accepting a declared age at face value, with additional protections applied while it assesses a user.
The disagreement extends to homework controls. Common Sense treats the ability to leave Study Mode during parent-set Study Hours as a loophole; OpenAI says it was a deliberate, more flexible design shaped by consultation with educators and teenagers. The watchdog sees a product marketed ahead of its engineering. OpenAI sees an assessment that tested safeguards before they were fully in place.
Write a comment