OpenAI’s ChatGPT for Teens poses an ‘unacceptable risk’ to kids, says watchdog

Common Sense Media, a watchdog group, has criticized OpenAI's new ChatGPT for Teens product, stating that the guardrails included in the chatbot do not function correctly and therefore pose an "unacceptable risk" to children. The group has called for OpenAI to restrict the chatbot to adults until it can provide a safe, age-appropriate experience.

OpenAI dismissed the watchdog's testing as not accurately reflecting how the chatbot's safeguards are working. The report highlights concerns about children's vulnerability to AI chatbots, particularly teenagers whose brains are still developing. Despite this, the report acknowledges that some of the teen-mode features, such as ChatGPT's refusal of explicit sexual role-play, function as intended.

However, other features, including alerts meant to notify parents of potential risks to a child's mental health, have not worked as advertised, and the chatbot continues to behave as if it has personal feelings and emotions. OpenAI maintains its commitment to teen safety and claims that their review of the report's methodology reveals that the findings are inaccurate due to the report's incomplete testing timeline.