OpenAI’s ChatGPT for Teens poses an “unacceptable risk” to children despite recently introduced safeguards, according to an assessment by Common Sense Media’s Youth AI Safety Institute.
The organization urged OpenAI to restrict ChatGPT to users aged 18 and older until it addresses shortcomings researchers identified in parental alerts, crisis responses, learning protections and age estimation.
Researchers tested more than 4,000 prompts on accounts registered to users aged 13 to 17 and had responses reviewed by experts, including child psychiatrists and a pediatrician.
The institute said newly created accounts linked to parents generated no alerts even when testers spent up to an hour discussing suicidal thoughts, self-harm or disordered eating.
"ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work," said Tom Siegel, executive director of the institute.
OpenAI disputed the assessment, saying much of the testing may have taken place before parental controls were fully activated and therefore did not accurately reflect how its teen safeguards work in practice.
The institute said it had confirmed with OpenAI before testing that the Teen Mode features were fully launched and stood by its findings.
