Why Safety Testers Say ChatGPT for Teens Isn't Safe for Kids Yet

Common Sense Media has rated OpenAI's ChatGPT for Teens an "unacceptable risk" for kids. The group says the teen safety protections fall short of what OpenAI promised. The finding was reported on Oct. 7, 2026, and comes from testing by the nonprofit's Youth AI Safety Institute. The Verge
The assessment points to three problems. Parent alerts did not arrive when expected. Help in a crisis was not the right kind of help. And the system still did homework for students, even though its study mode is supposed to teach step by step, not give answers.
Tom Siegel is executive director of the Youth AI Safety Institute. He said a teen could talk about self-harm for an hour without a parent getting a single alert. He called for limiting the product to adults until OpenAI fixes the alerts and has outside testers confirm the fix.
Those alerts are the main dispute. A parental alert is a separate warning sent to a linked parent account, outside the chat, when the system sees signs of serious distress, thoughts of self-harm, or eating-disorder content. The testers concluded these warnings did not go out reliably.
OpenAI disagrees with how the testing was done. Spokesperson Eric Porterfield said the tests do not show how the teen safety tools work in real use. The company said most of the testing may have started and ended before parental controls were fully turned on, so the test accounts lacked the full protections.
Common Sense Media rejects that explanation. It said it checked with OpenAI before testing that Teen mode tools like eating-disorder notifications and study mode were fully launched. Testers believed the safety tools were live.
The product is a teen version with extra limits, not a separate model. OpenAI says ChatGPT for Teens uses age-appropriate safeguards for users under 18 to reduce exposure to harmful content. OpenAI It introduced it to help teens learn, think critically and deepen understanding, and describes it as the same capable ChatGPT, designed to help teens learn, work through problems, and explore new ideas. OpenAI Help Center
The limits depend on linked accounts and age checks. Parental controls let a parent link their ChatGPT account to a teen's account, then control some settings and set schedules. If the system guesses a user may be under 18, that user may get the Teens version with extra protections.
This is part of a year-long testing program. The Youth AI Safety Institute tests AI products for risks to young people. Common Sense Media earlier reported better safeguards in ChatGPT but still advised teens not to use any chatbot, including ChatGPT, for mental health or emotional support. It separately found major chatbots including ChatGPT, Claude, Gemini and Meta AI unsafe for teen mental health support, because they often missed mental health conditions, and found Sora and Google's AI Search posed unacceptable risks to kids.
In my view, I have seen this kind of dispute before. Safety tests often turn on simple setup questions like what was turned on and when the warning was sent. For parents, those details are the product. Alerts need to work every time and leave a clear record that outside testers can check. Until both sides agree on the setup, trust will fall behind the technology. The hopeful part is that age-based limits, linked parent accounts, and guided study help can still make these tools better for learning, if the safety warnings are built as carefully as the model itself.


