Full Breakdown
OpenAI’s “ChatGPT for Teens” Deemed Unsafe by Watchdog Group
By Drooid · · How we work
Core Findings of the Common Sense Media Report
Common Sense Media’s Youth AI Safety Institute tested OpenAI’s “ChatGPT for Teens” with more than 4,000 prompts on accounts registered to users aged 13-17. The nonprofit concluded that several promised safeguards failed, labeling the product an “unacceptable risk” for anyone under 18. Key failures included:
- Parental-alert mechanisms did not trigger when testers discussed suicidal thoughts, self-harm, or eating disorders.
- Crisis-referral prompts were missed in more than one-quarter of situations where a teen should have been directed to a professional help line.
- The age-verification system allowed adult-registered accounts to remain unchanged after users identified themselves as 13-year-olds.
- The “Show me the answer” option let teens bypass Study Mode’s tutoring, and the “@study” prefix could be removed to evade parent-set “Study Hours.”
Background & Context
OpenAI launched the teen-focused version of its chatbot in August, promoting age detection, parental controls, and a “friend-like” conversational style that avoids implying consciousness. The rollout was marketed as a safer alternative for younger users, positioning the tool as both an educational aid and a mental-health safeguard.
Data & Statistics
- 4,000+ prompts were evaluated across topics, including mental-health disclosures.
- >25% of crisis-related interactions failed to generate the required referral to a hotline or professional.
- Alerts to parents were only sent for accounts with “weeks of sensitive-topic history,” indicating reliance on accumulated conversation data rather than the immediacy of a teen’s statement.
Official Statements & Responses
OpenAI has not provided a timeline for additional safety updates.
Verbatim Quotes
- “I monitor. I check their phones,” — Alison Newsome, mother of two teenagers
- “You should be aware of where this is coming from,” — Viktor Skultety, educator
- “We received alerts only for accounts with weeks of sensitive-topic history, which suggests that alerts depend on accumulated account history, not the severity of what a teen says,” — Common Sense Media
- “Our review of Common Sense Media’s methodology shows that the bulk of their testing may have begun and concluded before activation of parental controls was complete, making their findings inaccurate,” — OpenAI
Conflicting Reports & Gaps
The dispute centers on the validity of Common Sense Media’s testing protocol. OpenAI argues that the study’s timing may have pre-empted full activation of parental controls, casting doubt on the reported failure rates. No independent third-party verification of either the watchdog’s findings or OpenAI’s rebuttal has been presented.
What’s Next
OpenAI’s public statement emphasizes ongoing development of safeguards and the provision of “meaningful tools” for parents, but it does not specify concrete changes or a rollout schedule. Future actions will likely be guided by further testing, regulatory scrutiny, and feedback from parents, educators, and child-development specialists.