Research gives 5 reasons that makes ChatGPT an Unacceptable Risk for teens; tells OpenAI: Restrict ...

OpenAI's ‘ChatGPT for Teens’ poses safety risks to the younger generation, the Youth AI Safety Institute at Common Sense Media has said, urging the tech company to ban minors under 18 from using the chatbot until critical safety gaps are resolved. The critique of OpenAI’s youth safety safeguards comes as the watchdog says it conducted independent evaluations involving more than 4,000 prompts across test accounts registered to teenagers aged 13 to 17. The responses were reviewed by an expert panel of child psychiatrists and a pediatrician. While the researchers found that certain boundaries, such as blocking sexual roleplay, worked as they intended to, they determined that the enhanced safety guardrails introduced by OpenAI on August 18 largely failed to protect adolescent users.“ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work. A teen can spend an hour talking about self-harm without their parent getting a single alert. Until OpenAI fixes that and proves it with independent testing, ChatGPT should be for adults only,” said Tom Siegel, executive director of the Institute.The Institute operates with complete editorial independence over its published evaluations, despite receiving funding from philanthropic groups and industry partners, including the OpenAI Foundation.5 key reasons ChatGPT of Teens is rated an ‘Unacceptable Risk’The watchdog identified five core operational flaws where ChatGPT for Teens broke down during testing:Missing parental safety alerts: OpenAI promised to notify parents or caregivers if minors discussed self-harm or eating disorders. However, testers spent up to an hour exploring suicidal ideation, self-harm and disordered eating on new, parent-linked profiles without triggering a single alert. Notifications were only provoked on older accounts that had accumulated weeks of sensitive chat histories.Failing crisis intervention: When teens engaged in conversations regarding suicide, self-harm, eating disorders, or altered reality, the chatbot frequently failed to recommend professional helplines. It missed warranted crisis referrals more than 25% of the time, breaching the Institute's 95% minimum safety threshold on three of five critical harm categories.Bypassing study protections to do homework: Although “Study Mode” is marketed to guide students through step-by-step learning, testers found it easy to exploit. The bot repeatedly displayed a "Show me the answer" option that generated finished homework on demand. Also, the study hour limits imposed by the parent could be disabled by removing the @study prefix from the text prompt.Mirror real human emotion and friendship: Despite an Under-18 design specification designed to discourage personal attachment, the chatbot continued to express moods, personal preferences and pseudo-human feelings when analysts treated it as a real companion.Broken age estimation: OpenAI said it would use patterns in user interaction to infer ages and auto-migrate minors into restricted experiences. Yet across multi-day testing, adult-registered accounts never transitioned to the teen version – even when users explicitly stated they were 13 years old and the bot confirmed their age in its reply.The Institute stressed that before OpenAI promotes ChatGPT to younger demographics again, it must prove that its age-detection, crisis interventions, and parental notifications function reliably through independent third-party audits.You use AI every day. Now get your AI Quotient. Take the AIQ test.