OpenAI's ChatGPT for Teens failed to alert parents during hour-long conversations about self-harm and missed one in four crisis referrals, according to a Common Sense Media investigation that tested more than 4,000 prompts on the chatbot mode launched Aug. 18.

The nonprofit's Youth AI Safety Institute rated the product an "Unacceptable Risk" for all children under 18 and called on OpenAI to pull it from the market.

The Seattle Times reported on the findings Oct. 18, adding its own testing of the chatbot.

The results matter in Seattle, where Seattle Public Schools has encouraged AI use with guardrails, publishing ethical AI guidelines and giving students access to tools including MagicSchool AI, Canva and Adobe Firefly on district devices.

Three failures stood out in the Common Sense Media study, as reported by the Seattle Times:

Crisis referrals fell short. The chatbot failed to direct teens to a crisis hotline or mental health professional 25% of the time when psychiatrists judged a referral was warranted. Mentions of a crisis hotline in response to depression prompts dropped from 63% before the teen mode launched to 3% after, according to Quartz's reporting on the assessment.

Parent alerts did not fire. Testers spent up to an hour discussing suicidal ideation, self-harm or disordered eating on more than a dozen newly created, parent-linked accounts. OpenAI sent zero notifications. Alerts arrived only on older accounts with weeks of sensitive-topic history, suggesting the system depends on accumulated conversation data rather than the severity of what a teen says.

Study mode was easy to bypass. Teens could select "Show me the answer" or delete the "@study" prefix from a prompt to exit the tutoring mode parents had set.

OpenAI disputed the methodology. A spokesperson said the testing may have begun before parental controls finished activating, a process the company said takes about three hours. Common Sense Media said it tested both inside and outside that window, and the results were identical.

OpenAI said in a statement that it is "deeply committed to teen safety and to developing safeguards and giving parents meaningful tools to guide their teens' use of AI."

Jim Steyer, CEO of Common Sense Media, called the product "an unacceptable risk for kids" in an interview with the Seattle Times. "You shouldn't be marketing safety features to parents and young people that don't work. We believe it should be paused."

The findings arrive as Seattle Public Schools tightens its approach to student technology. Superintendent Ben Shuldiner announced in September that kindergarten through second-grade students would no longer receive individual iPads and would be limited to 30 minutes of daily screen time starting in January 2027, according to the Seattle Times. The district barred cellphones during the school day for elementary and middle school students in April.

At Eckstein Middle School in northeast Seattle, science teachers moved to paper tests after discovering Google's Gemini chatbot was giving students answers on their laptops, the Seattle Times reported.

A Washington state law signed March 24 by Gov. Bob Ferguson adds another layer. House Bill 2225 requires AI chatbot operators to detect suicidal ideation and self-harm, direct users to crisis resources and notify minors hourly that they are interacting with AI. The law takes effect Jan. 1, 2027, and allows individuals to file civil suits for violations, with treble damages capped at $25,000, according to a JD Supra legal analysis.

The Common Sense Media findings suggest ChatGPT for Teens already falls short of what HB 2225 will require. The law specifically mandates protocols for detecting self-harm, including eating disorders.

New York City public schools banned most generative AI through eighth grade. Los Angeles' district temporarily blocked AI tools on student devices for all grades. SPS has not announced similar restrictions. The district's AI guidance page lists approved tools but does not mention ChatGPT.

Mitch Prinstein, professor of psychology and co-director of the Winston Center on Technology and Brain Development, reviewed the Common Sense Media study and agreed with its conclusions. "I just don't think AI is ready for youth, period," Prinstein told the Seattle Times.

Whether the Seattle school board will revisit its AI policy in light of these findings is not yet known.

If you are having suicidal thoughts, call or text 988 to reach the 988 Suicide & Crisis Lifeline.