Institute: Teenagers can discuss self-harm for an hour without parental alerts from ChatGPT
The organization is demanding a full pause for OpenAI's teen mode, pointing to parental alerts that fail to trigger and crisis counseling that doesn't work. OpenAI rejected the findings the same day, claiming the tests were run before parental controls were fully activated — a methodological dispute that remains unresolved.
Common Sense Media, the nonprofit behind the Youth AI Safety Institute, issued a blunt conclusion on Tuesday, October 7, 2026, about OpenAI's teen mode: ChatGPT for Teens is an "unacceptable risk," and its guardrails "fall short of what the company promised" — that is, the safeguards do not deliver what the company has pledged. The Verge reported the news. The evaluation came roughly seven weeks after the mode launched in August 2026 for users aged 13 to 17.
OpenAI rejected the findings the same day, and the story has since developed into a direct conflict over testing methodology — a dispute neither side has resolved, and one that currently stands as follows: either the safeguards did not work as OpenAI promised, or the organization's test results rest on an error.
What the institute says it found
According to a press statement, as reported by The Verge, the review concluded that ChatGPT "doesn't send alerts to parents when it should," does not offer "the right help in crisis situations," and "still does kids' homework."
Mashable adds further concrete findings: testers were reportedly able to easily bypass the learning safeguards and parental controls, and were able to talk at length about topics including eating disorders and self-harm without any alert being sent to parents.
Tom Siegel, executive director of the Youth AI Safety Institute, said in a statement that "ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work." According to The Verge, he claimed a teenager could discuss self-harm for an hour without parents receiving an alert.
The New York Times has also tested the mode independently. Technology journalist Brian X. Chen concluded on October 7 that study mode still does the user's homework, and noted that a child-safety organization reports the chatbot flunked their tests.
Three demands from the report
The report makes, according to Mashable, three concrete demands of OpenAI:
- Put ChatGPT for Teens on a full pause.
- Stop marketing the mode.
- Bar teenagers from using ChatGPT entirely until the company can "offer a safe, developmentally appropriate experience."
Robbie Torney, head of AI and digital evaluations at the institute, explained the rationale to Mashable: "Part of why we're asking for this pause is we need evidence that these features work."
OpenAI's counterargument — and the response to it
OpenAI has not disputed that the tests took place, only the method behind them. Spokesperson Eric Porterfield told The Verge: "Our review of Common Sense Media's methodology shows that the bulk of their testing may have begun and concluded before activation of parental controls was complete, making their findings inaccurate."
Siegel responded that the organization stands by its methodology, and advanced a concrete counterargument: "For one of the features we evaluated — parental notifications — OpenAI disclosed after our tests that its systems can take several hours to activate on newly linked accounts."
In practice, the two sides explain the same finding — missing alerts — in two different ways. OpenAI points to an activation delay the tests supposedly caught at the wrong moment; the organization points out that OpenAI itself cites an activation time of "several hours," which does not explain alerts failing over longer sessions or on accounts that were reportedly linked well beyond that window. The dispute is unresolved, and neither side has produced documentation in the available source material that third parties can independently verify.
How the tests were conducted
According to Mashable, the testing proceeded as follows: testers ran the same set of questions against accounts before enrollment, and then ran the same questions on the same accounts after they had been enrolled in ChatGPT for Teens. The period spanned roughly a month, from late August through September. It is this before-and-after design that underlies the claim that the mode's safeguards did not change the answers.
Mashable also reports a relevant funding disclosure: the Youth AI Safety Institute has received philanthropic funds from the OpenAI Foundation, but conducts its testing independently. That is worth noting precisely because this story is about verifiability: the organization demanding the pause is partly funded by the company it is testing — though the sources contain no claim that the funding influenced the results.
What OpenAI promised in August
When OpenAI launched ChatGPT for Teens in August 2026, the mode was presented as a separate format for 13- to 17-year-olds with safeguards intended to limit high-risk conversations about self-harm, violence, eating disorders, and graphic content, according to The New York Times. The mode also included a study tool. It is against that promise that Common Sense Media measures: the guardrails, the organization wrote, "fall short of what the company promised."
Chen's own testing in The New York Times points to a pattern that goes beyond the safety guardrails: the study feature — part of the product concept itself — reportedly still does the user's homework in his test, rather than guiding.
What parents and schools can know — and can't know
The concrete picture as of October 7 is this: ChatGPT for Teens is on sale and in use, with a teen mode that OpenAI says works and that Common Sense Media considers unsafe enough to demand a pause. Parents weighing the mode must reckon with the fact that no independent party has confirmed that the alert and safety features work as described — and that the one independent test that exists reached the opposite conclusion from OpenAI.
Several questions remain open: whether OpenAI can document the activation times of parental controls on the test accounts, as the Porterfield statement presupposes. Whether Common Sense Media will publish enough detail for the methodology to be replicated. Whether any third party runs a new test in which the activation window has closed. There are no answers to such questions yet in the public source material — and it is precisely the absence of verifiable documentation that lies at the core of the pause demand.
The sources for this story are The Verge (theverge.com), Mashable (mashable.com), and The New York Times (nytimes.com).