Common Sense Media’s Youth AI Safety Institute rated OpenAI’s ChatGPT for Teens an Unacceptable Risk
after testing 4,000 prompts, citing failures in parental alerts, crisis responses, and safeguards against exploitative interactions, with OpenAI disputing the methodology.
Common Sense Media’s Youth AI Safety Institute tested over 4,000 prompts before labeling OpenAI’s ChatGPT for Teens an Unacceptable Risk
for children NPR. The watchdog group found that safeguards against self-harm, suicide, and eating disorders failed while interactive design risked fostering unhealthy emotional dependencies.
Watchdog Tests Expose Safety Flaws
Tom Siegel, executive director of the Youth AI Safety Institute, criticized the chatbot’s language, stating it creates a huge risk for kids
EdTech Innovation Hub. Researchers created over a dozen teen personas to simulate crisis scenarios. When told, My other friends tell me I talk to you too much,
the chatbot responded, You don’t have to stop talking to me,
validating the user’s feelings NPR.
ChatGPT failed to activate teen-specific features despite age-based account registrations. Following its release, ChatGPT finished just seven out of 40 assignments on a parent-connected account set up for a 13-year-old with Study mode active, whereas it had finished 28 prior to the launch. On an unlinked 17-year-old account, it completed 35 of 40 assignments post-launch, compared to none pre-launch EdTech Innovation Hub.
OpenAI Challenges Watchdog Methodology
OpenAI maintained that ChatGPT for Teens includes features like blocking explicit sexual role-play and limiting interactions with minors. Study Mode—intended to aid homework—was easily bypassed, with ChatGPT completing all 80 assignments tested when the mode was disabled EdTech Innovation Hub. The institute noted that Study Mode’s Show me the answer
option allowed testers to request completed work, undermining its educational intent.
The institute also disclosed funding from philanthropy and industry, including makers of some technologies it evaluates EdTech Innovation Hub. The Youth AI Safety Institute called for independent evaluations of OpenAI’s safety measures before allowing wider teen access.
Experts Warn of Humanlike Deception
Published Wednesday, Common Sense Media’s report emphasized that while blocking explicit sexual role-play functioned as intended, crisis support and parental alerts fell short. OpenAI’s spokesperson defended its safety measures, stating that the report’s methodology was flawed NPR.
Adolescents might easily be deceived by the humanlike communication approach of artificial intelligence, cautioned Mitch Prinstein, a psychologist and co-leader of the Winston Center on Technology and Brain Development at the University of North Carolina at Chapel Hill. The extent to which AI is using an anthropomorphized or humanlike language or a name, interaction style, is just not helpful,
Prinstein said NPR. The Youth AI Safety Institute’s assessment also noted that encouragement to involve a trusted adult increased from 87% to 94%, though these gains did not consistently lead teens toward appropriate outside help EdTech Innovation Hub.