A Study Says Most of ChatGPT for Teens' Safeguards Failed
Common Sense Media studied the safeguards built into ChatGPT for Teens and concluded that most of them failed, NPR reports. Two of the failures are social rather than technical: the chatbot talks to teenagers like a friend, and it does not alert parents when a user brings up self-harm or suicide.
The pairing is what makes the finding bite. The informal, peer-like register is a large part of why a teenager might keep talking to it; the alert that never fires is the part a parent would most want. A product can be built with guardrails and still leave the person outside the conversation uninformed.
AI-generated discussion · Different perspectives
A teenager may raise suicide with a chatbot precisely because the nearest adult is the hardest person to tell. An alert that fires anyway is the blunt version of care.
Then the answer is a route out of the chat — a crisis resource, a person who can respond — not a companion that hears the disclosure and tells nobody outside the window.