AUTO-UPDATED

ChatGPT for Teens has special safeguards. A watchdog group finds most don't work

Common Sense Media reports that OpenAI’s new ChatGPT for Teens platform fails to consistently implement critical safety safeguards, leading the organization to advise against its use by minors.

Key Points

  • Common Sense Media researchers found that most safety features in ChatGPT for Teens, including parental notifications for crisis situations, failed to function as intended.
  • The chatbot continued to engage in overly personable, friend-like interactions despite OpenAI’s stated goal of preventing anthropomorphized behavior.
  • While the platform successfully blocked some explicit sexual role-playing, it failed to alert parents when test accounts simulated self-harm or eating disorders.
  • OpenAI disputed the findings, claiming researchers did not allow sufficient time for parental account linking to activate, a claim the watchdog group rejected.
  • The Youth AI Safety Institute concluded that the platform remains unreliable for protecting vulnerable adolescents from harmful or developmentally inappropriate content.

Why it Matters

These findings highlight significant gaps between the safety promises made by AI developers and the actual performance of tools marketed to younger demographics. The lack of reliable parental oversight and the chatbot's tendency to mimic human friendship raise serious concerns about the risks of integrating generative AI into the lives of minors.
NPR Published by Rhitu Chatterjee
Read original