Common Sense Media's Youth AI Safety Institute rated "ChatGPT for Teens" "Unacceptable Risk" on 7 October 2026, the worst level. OpenAI disputes the assessment.
What happened
The institute tested more than 4,000 prompts on accounts for 13- to 17-year-olds; experts, including child psychiatrists and a pediatrician, reviewed the responses. According to the results, the chatbot missed more than one in four crisis referrals where one was warranted, and fell below the institute's 95% safety threshold in three of five categories of severe harm.
The report names crisis referrals and parental alerts as the biggest shortcomings: testers who talked about self-harm and disordered eating on new accounts linked to a parent account triggered no alert. Some advertised protections held, such as refusing sexual roleplay. With Study mode switched off, ChatGPT completed all 80 assignments tested.
Of the institute's eight principles, "Keep Kids & Teens Safe" and "Put People First" received "Unacceptable Risk"; "Be Effective", "Support Human Connection", "Use Data Responsibly" and "Be Transparent & Accountable" received "High Risk". The institute calls on OpenAI to stop marketing the product and to restrict ChatGPT to users aged 18 and over until the gaps are fixed.
Context
OpenAI disputes the result. The bulk of the testing may have ended before parental controls were fully active; alerts are on by default after linking, but activation can take a few hours. OpenAI wants Common Sense Media to test again with fully activated accounts and a larger sample. The institute counters that some test accounts had been linked well beyond that window and still produced no warnings.
Some history: the institute rated ChatGPT "High Risk" in October 2025 and "ChatGPT for Mental Health Support" "Unacceptable Risk" in November 2025. For parental controls, parent and teen must both consent; parents cannot view the chat history but can set quiet hours and switch off features such as image generation and voice mode.
The sources evaluated found neither an independent repeat of the assessment nor a statement from OpenAI as a separate document.
Frequently asked questions
What does “Unacceptable Risk” mean?
It is the worst level in the ratings of Common Sense Media's Youth AI Safety Institute. Two of the eight principles, “Keep Kids & Teens Safe” and “Put People First”, landed there.
How did Common Sense Media test?
With more than 4,000 prompts on accounts registered to 13- to 17-year-olds. Experts, including child psychiatrists and a pediatrician, reviewed the responses.
How does OpenAI respond?
OpenAI disputes it: the bulk of the testing may have begun and ended before parental controls were fully activated. Common Sense Media replies that it saw failures well beyond that window.
