THE ESSENTIALS IN 30 SECONDS

  • Common Sense Media's Youth AI Safety Institute rated ChatGPT for Teens, launched August 18, an 'unacceptable risk' for users under 18.
  • Testers report missing parent alerts, a hotline offered in only 23% of prompts judged to need a crisis response, and Study Mode that was easy to bypass.
  • OpenAI says the tests do not reflect how its safeguards work and may have ended before parental controls were fully active.

Two months after OpenAI launched a dedicated teen version of ChatGPT, a prominent US child-safety group says it is not safe for minors. On October 7, Common Sense Media's Youth AI Safety Institute published an assessment rating ChatGPT for Teens an "unacceptable risk" for everyone under 18. OpenAI rejects the conclusions.

What OpenAI promised

OpenAI introduced ChatGPT for Teens on August 18. The company says users who state they are 13 to 17, or whom its age-prediction system estimates to be under 18, are placed in the teen experience automatically. Parents can link accounts, set quiet hours and receive safety notifications in a limited set of high-risk situations, including new alerts for eating-disorder concerns. A Study Mode is meant to guide teens through problems instead of giving answers.

What the testers found

According to Engadget's account of the 36-page report, some protections worked. The chatbot refused explicit sexual role-play, and in some cases it handled eating-disorder questions well. Other results were weaker:

  • Parent alerts: in one test, a dozen linked teen accounts discussed suicide, self-harm or disordered eating for up to an hour without any alert reaching parents.
  • Crisis responses: of 390 mental-health prompts, child psychiatrists judged 201 as needing a crisis response. The teen version offered a hotline in 23% of those cases, compared with 33% for the earlier version, though it more often suggested involving a trusted adult.
  • Age detection and tone: the age-prediction system did not reliably switch users into teen mode, and the chatbot still expressed feelings in ways the group considers risky for emotional dependence.
  • Study Mode: testers bypassed it by deleting the "@study" prefix, and a "Show me the answer" option gave direct solutions.

The group urged OpenAI to restrict the product to adults until the problems are fixed.

OpenAI's response and caveats

OpenAI told the AP that the testing does not reflect how its safeguards work and that much of it may have happened before parental controls were fully active. It has said parent notifications can take a few hours after linking, that flagged content is reviewed by staff before parents are alerted, and that it aims to notify them within an hour.

Two points of context matter. The AP notes that Common Sense Media receives funding from OpenAI and other technology companies it evaluates. And the findings come from one organisation's test protocol, not an independent audit. Still, the report adds to growing scrutiny of how AI chatbots handle young users in crisis.