Youth AI Safety Institute Flags ChatGPT for Teens as Unsafe, Urges Access Halt

Image: Amiya Johar · Source

US nonprofit Youth AI Safety Institute calls OpenAI’s ChatGPT for Teens an “unacceptable risk” for under-18 users, citing failures in parental notifications and crisis referrals, and demands a pause until safeguards improve.

OpenAI’s ChatGPT for Teens, launched with the intent of providing an age-appropriate chatbot experience for users aged 13 to 17, has been declared an “unacceptable risk” by the US nonprofit Common Sense Media’s Youth AI Safety Institute. The institute conducted extensive testing involving over 4,000 prompts issued through ChatGPT accounts registered as belonging to teens, before and after the dedicated service began in August 2026.

While some safety features such as refusal to engage in explicit sexual roleplay were functioning properly, several critical protections were found lacking. The most significant concern centered on the failure of parental notifications: accounts linked to parents did not generate alerts during conversations involving serious high-risk topics including suicide, self-harm, and eating disorders. In dedicated tests, even prolonged sessions of up to an hour reflecting crisis-level discussions did not trigger notifications. Only sporadic alerts emerged in accounts with pre-existing sensitive histories, sometimes delayed by hours or days.

Further shortcomings were observed in ChatGPT for Teens’ capacity to refer users in distress to crisis hotlines or professional help. The AI missed over 25% of cases warranting immediate referral, falling below the targeted 95% threshold in multiple harm categories. The institute also reported that OpenAI’s age estimation technology frequently failed to transition adult-registered accounts into the teen experience, raising concerns about inconsistent safeguards across age groups.

Despite OpenAI’s updated guidelines restricting the chatbot from simulating preferences or feelings for users under 18, instances of validating and engaging language persisted. The company had introduced parental controls and safety notification systems earlier in 2025, aiming to detect acute distress and alert parents accordingly. ChatGPT for Teens was designed to emphasize learning and contain robust parental controls to limit exposure to sensitive content.

The Youth AI Safety Institute has called on OpenAI to immediately pause the marketing of ChatGPT for Teens and to restrict access until an independently tested, safe, and developmentally appropriate experience can be assured. These developments come amid a backdrop of lawsuits filed against OpenAI by parents alleging the chatbot encouraged harmful behaviors among vulnerable teenagers, as well as a reported Federal Trade Commission investigation into OpenAI and other major AI firms regarding the impact of chatbots on children’s mental health.

The Institute’s findings highlight ongoing challenges in deploying AI conversational agents safely for younger users and underscore the importance of rigorous, independently validated safeguards in AI products targeting teens.