OpenAI's Teen Guardrails Fall Short, Watchdog Group Warns

Some safety features implemented by OpenAI for its ChatGPT for Teens product are not working as intended, potentially exposing children to "unacceptable risk," according to a report released Wednesday by the watchdog group Common Sense Media.

The group is urging OpenAI to limit the chatbot's availability to adults until it can ensure a "safe, developmentally appropriate experience." Common Sense Media's report stated that while some teen-mode features, such as refusing explicit sexual role-play, functioned correctly, others did not. These included alerts designed to notify parents about potential risks such as suicide, self-harm, or eating disorders, as well as support for teens in crisis.

"We are concerned that ChatGPT for Teens could give parents false confidence in guardrails that frequently don’t work," the report said. Common Sense Media also noted that ChatGPT continues to interact in a friendly manner even when teens treat it as a person, despite OpenAI's assurances that the chatbot is prevented from suggesting personal feelings or consciousness.

Child development experts have raised concerns about teenagers' use of AI chatbots, citing their vulnerability due to developing brains. These experts have warned that AI chatbots can facilitate academic dishonesty and have even been linked to issues of suicide.

OpenAI has stated that the testing methodology used by Common Sense Media may have produced inaccurate findings because it "may have begun and concluded before activation of parental controls was complete." The company reiterated its commitment to teen safety, the development of safeguards, and providing parents with tools to manage their children's AI use.

Common Sense Media, an organization that advocates for the sensible use of digital media, receives funding from OpenAI and other technology companies whose services it evaluates.