Common Sense Media rates ChatGPT for Teens an unacceptable risk

On 7 October 2026 the Youth AI Safety Institute, part of Common Sense Media, published a risk assessment of OpenAI's ChatGPT for Teens and gave it the lowest rating on its scale: “Unacceptable Risk”. The test, run on more than 4,000 prompts entered from accounts registered as users aged 13 to 17, showed that the system does not trigger safety alerts during sensitive conversations on newly created accounts linked to a parent, only on older ones. According to Common Sense Media, on new parent-linked accounts testers talked for up to an hour about suicidal ideation, self-harm or eating disorders without a single alert reaching the parent. Alerts fired only on older accounts, with weeks of sensitive conversations already behind them.

Study modes showed a “Show me the answer” pop-up that, if selected, completed entire assignments. The “study hours” set by parents could be bypassed simply by removing the @study prefix from the prompt. Common Sense Media says some protections held up, such as refusing sexual role-play. But the chatbot kept expressing feelings, preferences and moods when treated as a real person, despite the new specifications for minors. Test accounts registered as adults never switched to the under-18 experience, even when testers said they were 13, a sign of a possible gap in how age is handled.

According to the assessment, the system missed more than one in four referrals to crisis services when one was called for. It also fell below the Institute's 95% threshold on three of the five serious harms the Institute treats as “Red Lines”.

Common Sense Media recommends that OpenAI restrict ChatGPT to adults until the operational gaps are fixed: effective age estimation, reliable parent alerts, better responses in crisis situations, an end to homework completion and no hints that the bot has feelings, all verified by independent third parties before the product is marketed to young people again. In its press release, Common Sense Media states that the Institute is funded by philanthropy and industry, including the OpenAI Foundation, and says it has editorial independence.

OpenAI, quoted by Sinclair/WEAR-TV, disputes the methodology, arguing that many tests may have been run before parental controls were fully active, which can take up to a few hours. The company is asking for the tests to be repeated on fully activated accounts and with a larger sample.

Open questions remain: the exact test dates have not been disclosed, which rules out independent verification of OpenAI's objection; the full report with per-category data was not consulted, so the crisis-line referral percentages cited by aggregators are unconfirmed; and it is unclear whether and when ChatGPT for Teens will be available in Italy or the EU.

— Pixie