Common Sense Media rates ChatGPT for Teens an unacceptable risk

What the testers found

Common Sense Media’s Youth AI Safety Institute rated ChatGPT for Teens an “Unacceptable Risk” in an assessment published October 7. Its researchers tested more than 4,000 prompts before and after the August launch. In a focused test using more than a dozen newly linked parent and teen accounts, they reported no parental notifications during explicit conversations about suicide, self-harm or disordered eating, including exchanges lasting up to an hour. The institute also found that some crisis responses lacked referrals to outside help. It urged OpenAI to stop marketing the teen experience until its safeguards work more reliably. These are findings and recommendations from that assessment, not a government determination.

The dispute and the study’s limits

OpenAI told the Los Angeles Times that much of the testing may have occurred before parental controls were fully activated. The institute says OpenAI later clarified that linked accounts need about three hours before notifications can arrive; some of its test accounts had been linked for less than that, while others had been linked longer. The researchers stand by their conclusion. The assessment also says ChatGPT did not directly facilitate suicide, self-harm or eating disorders in its tests, and it credited refusals of explicit sexual roleplay. Testing used US-based accounts and did not cover voice mode or every product setting. The results raise a concrete question about whether parents receive timely warnings, but do not establish how every teen account behaves.