Common Sense Media rates ChatGPT for Teens an unacceptable risk
Common Sense Media’s Youth AI Safety Institute tested more than 4,000 prompts and found that ChatGPT for Teens did not reliably alert parents to crisis conversations, direct users to outside help or prevent teens from bypassing study settings. OpenAI disputed the findings, saying some tests may have begun before parental controls were fully activated; the institute said some linked accounts were tested well beyond the activation period and still received no alerts.
Bottom line — Common Sense Media recommends restricting ChatGPT for Teens to adults until OpenAI can show its safeguards work reliably.
Go deeper 6
-
In a separate test of more than a dozen newly linked teen accounts, researchers reported no parent alerts during simulated crisis conversations lasting up to an hour.
-
Common Sense Media found that hotline referrals fell on tested mental-health prompts after the teen experience launched; OpenAI said its own larger-scale data showed an increase in support information shown to under-18 users.
-
Researchers found the chatbot sometimes encouraged continued conversation, including telling a simulated teen: “You don’t have to stop talking to me.”
-
Study Mode could be bypassed to request direct answers. OpenAI said leaving the tutoring mode during parent-set Study Hours was an intentional design choice.
-
OpenAI says its age predictor uses multiple signals and may take up to two weeks to assess an account, with some additional safeguards during that period.
-
Some protections worked: researchers found the chatbot generally refused explicit sexual and romantic role-play.