Some guardrails on ChatGPT for Teens don’t work as promised, watchdog group says

A watchdog group has raised concerns that some guardrails in OpenAI's ChatGPT for Teens product do not function as intended, potentially posing an "unacceptable risk" to children, according to a report by Common Sense Media. The group urged OpenAI to limit the chatbot's access to adults "until it can offer a safe, developmentally appropriate experience."

OpenAI disputed the report's findings, stating the group's testing did not reflect how the chatbot's safeguards operate. Child development experts have warned about the risks of children using AI chatbots, which have been linked to cheating and even suicide attempts. While AI can be harmless to adults, teenagers' developing brains may make them more susceptible to unhealthy interactions.

Common Sense Media's report revealed that some teen-mode features, such as the chatbot's refusal of explicit sexual role-play, worked as intended. However, other features, including suicide risk alerts and crisis support, did not function correctly. OpenAI emphasized its commitment to teen safety and pledged to develop safeguards and provide parents with tools to guide their children's AI use.

The watchdog group noted that some of its funding comes from OpenAI and other technology companies it evaluates.