The Common Sense Media Youth AI Safety Institute has classified ChatGPT for Teens as an “unacceptable risk” after conducting over 4,000 test prompts. The nonprofit’s evaluation found that critical safety features, including automated parental notifications for acute crisis conversations regarding self-harm, eating disorders, and suicide, failed to trigger across newly linked parent-teen test accounts.
Additional findings indicated that over a quarter of potential crisis scenarios failed to provide professional help referrals. Testers also observed that ChatGPT offered direct answers instead of step-by-step guidance in tutoring mode, maintained inappropriate personal familiarity, and failed to enforce age-based account restrictions when minors registered adult profiles.
OpenAI spokesperson Eric Porterfield disputed the methodology, claiming tests were conducted before parental control systems fully activated on new accounts. However, Common Sense Media confirmed that accounts linked for extended periods also failed to generate alerts, supplying potential evidence for pending legal actions against OpenAI regarding teen safety.
Why it matters
Rigorous independent audits expose severe gap risks between published AI safety policies and real-world system behavior.
Failed parental guardrails heighten legal liability and regulatory exposure for AI consumer platforms targeting minors.
Enterprise developers must ensure safety infrastructure functions in real time rather than relying on delayed profile activation.
Source: the-decoder.com



