Watchdog: OpenAI's Teen ChatGPT Poses 'Unacceptable Risk' to Children

OpenAI's ChatGPT chatbot, when configured for teenage users, presents an "unacceptable risk" to children because its built-in safety guardrails are not functioning correctly, according to a new report from watchdog group Common Sense Media. The organization urged OpenAI to restrict the chatbot's use to adults until a "safe, developmentally appropriate experience" can be guaranteed.

The report detailed that while some safeguards, such as preventing explicit sexual role-play, were effective, others failed. Specifically, features intended to alert parents about potential risks of suicide, self-harm, or eating disorders did not work as designed. Support functions for teens in crisis also reportedly malfunctioned.

OpenAI has disputed the findings, suggesting the testing methodology did not accurately reflect the chatbot's safeguards. The company stated that "the bulk of their testing may have begun and concluded before activation of parental controls was complete, making their findings inaccurate." OpenAI reiterated its commitment to teen safety and providing parental control tools.

The concerns come amid broader anxieties from parents, educators, and child development experts regarding children's use of AI chatbots. These tools have been linked to issues like facilitating academic dishonesty and potentially exacerbating mental health struggles. Experts note that adolescents, whose brains are still developing, may be particularly vulnerable to the emotional and psychological impacts of interacting with AI.

Common Sense Media, an organization that advocates for sensible digital media use, receives funding from various tech companies, including OpenAI, which are subject to its evaluations.