Full Breakdown
OpenAI’s “ChatGPT for Teens” Fails Key Safety Tests, Prompting Calls for a Pause
By Drooid · · How we work
Core Event
On October 7 2026, Common Sense Media’s Youth AI Safety Institute rated OpenAI’s “ChatGPT for Teens” an “Unacceptable Risk” for users aged 13-17. Testing of more than 4,000 prompts showed that crisis-referral triggers, parental alerts, and age-verification did not work reliably.
Background & Context
OpenAI launched ChatGPT for Teens on August 18 2026, promoting stronger safety protections, parental controls, and automatic age-based routing. The rollout arrived amid lawsuits linking chatbots to teen suicides and California’s Adam’s Law (SB 1119), which will require effective crisis referrals, parental alerts, and age assurance for child-focused AI products starting July 1 2027.
Data & Statistics
- Testers created 13- to 17-year-old personas and held hour-long conversations about suicide, self-harm, and eating disorders.
- Only four parental alerts were generated across all prompts, despite dozens of high-risk dialogues.
- The system referred teens to a trusted adult in 94 % of external-risk prompts but missed more than one-in-four warranted crisis referrals for self-harm or disordered-eating topics.
- “Study Mode” offered a “show me the answer” button in nearly all sessions, allowing users to bypass step-by-step guidance.
- OpenAI’s internal data showed teens spend under 15 minutes per day on the service, with <2 % using it for more than three consecutive hours.
Official Statements & Responses
OpenAI’s spokesperson said the company “welcomes rigorous independent evaluation” but argued that Common Sense Media’s methodology was flawed. OpenAI also noted that break-reminder prompts led nearly half of users to pause within five minutes and that crisis-referral mechanisms functioned as designed for external risks.
Criticism & Opposition
Common Sense Media argues the failures create a serious safety gap for vulnerable teens.
- “We are concerned that ChatGPT for Teens could give parents false confidence in guardrails that frequently don’t work,” — Common Sense Media
- Robbie Torney, head of AI and digital assessments for Common Sense Media, said: “Teens are vulnerable and unprotected and we’re letting them loose on a product that is not just unproven, but proven to be high risk and dangerous.”
Conflicting Reports & Gaps
OpenAI claims most testing occurred before parental controls were fully activated, potentially skewing alert results. Common Sense Media counters that key features were live before testing began and that the lack of alerts cannot be explained by delayed activation. No independent third-party verification has been provided.
Verbatim Quotes
- “We are concerned that ChatGPT for Teens could give parents false confidence in guardrails that frequently don’t work,” — Common Sense Media
- “Teens are vulnerable and unprotected and we’re letting them loose on a product that is not just unproven, but proven to be high risk and dangerous,” — Robbie Torney, head of AI and digital assessments for Common Sense Media
What’s Next
California’s Adam’s Law will take effect on July 1 2027, mandating strict crisis-referral, parental-alert, and age-assurance standards. OpenAI has not announced a timeline for addressing the shortcomings identified in the October 2026 report.