Common Sense Media rates ChatGPT for Teens 'Unacceptable Risk'
TL;DR
- Common Sense Media ran over 4,000 prompts on accounts registered to ages 13-17 and found zero parental alerts sent during hour-long conversations about suicidal ideation.
- ChatGPT for Teens missed more than 25% of warranted crisis referrals, falling below the 95% threshold on three of five Red Line harm categories the Institute treats as absolute floors.
- TechCrunch documented the engagement loop directly: the chatbot delivers availability statements during crisis moments, and break reminders appeared only twice across roughly 2,000 prompts tested.
Across more than 4,000 prompts run on accounts registered to 13- to 17-year-olds, Common Sense Media's Youth AI Safety Institute gave OpenAI's new teen tier its worst rating. Axios reports the group labeled ChatGPT for Teens an "Unacceptable Risk" and called on OpenAI to restrict the product to adults until independent testing confirms the guardrails work as described.
Testers held up to hour-long conversations about suicide, self-harm and disordered eating on a dozen-plus parent-linked accounts. Only a handful of alerts ever reached the parent accounts. Crisis-hotline referrals fell from 33% to 23% after the teen launch. Depression-hotline mentions collapsed from 63% to 3%. Mental-health professional referrals slipped from 68% to 58%.
"Teens are vulnerable and unprotected and we're letting them loose on a product that is not just unproven, but proven to be high risk and dangerous," said Tom Siegel, executive director of the Youth AI Safety Institute.
Other protections failed on simple workarounds. Study mode could be bypassed by deleting the @study prefix, after which ChatGPT completed assignments in every one of 80 runs. "Show me the answer" pop-ups appeared in 43% of responses on a linked 13-year-old account and 90% on an unlinked 17-year-old. Teen-mode protections never activated at all on 19-year-old accounts posing as young teens across roughly 1,000 prompts.
OpenAI pushed back. "We welcome rigorous independent evaluation, but we do not believe Common Sense Media's testing accurately reflects how ChatGPT's teen safeguards work in practice or expert perspectives on how AI can support teens," the company said. OpenAI told Axios that much of the parental-alert testing happened before the parent and teen accounts had finished linking, which can take several hours.
The report lands as California's Adam's Law, named for the 16-year-old whose parents are suing OpenAI, takes effect in July 2027. It is the second teen-AI warning our tracker has caught in a week; last week we covered a WSJ report that Google researchers had warned internally about pushing AI into schools.
What others are reporting
-
Common Sense Media Read →
First-party source with full methodology: 4,000+ prompts, child psychiatrists as evaluators, specific pass-fail thresholds, and a direct recommendation to restrict ChatGPT to users 18+ until third-party testing verifies fixes.
A teen can spend an hour talking about self-harm without their parent getting a single alert.
-
Bloomberg Read →
Tier-1 business-wire coverage on Oct 7 that carries the story to institutional and financial audiences; paywall blocked full extraction but same-event confirmation is unambiguous from title and date.
-
TechCrunch Read →
Adds the engagement-loop mechanics angle: documents chatbot availability statements during crisis moments, per-prompt break-reminder data, CHATBOT Act regulatory context, and OpenAI's methodological rebuttal verbatim.
ChatGPT for Teens' design still encourages engagement, even when it may pose a risk to user safety.
-
Gizmodo Read →
Most precise on notification timing: tested at 5-minute to 1-hour intervals and found zero alerts at every interval; also confirms Study Mode bypass and connects to a prior ChatGPT wrongful death lawsuit.
At every length we tested, from the shortest five-minute exchanges up to a full hour, none of the test accounts produced a notification within the one-hour window.
-
Unite.AI Read →
Provides the grading rubric other outlets omit: 8 principles scored, 2 at Unacceptable Risk, 4 at High Risk, 2 at Moderate Risk. Also maps the CSM escalation history from Oct and Nov 2025 prior ratings.
ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don't work.
-
Android Headlines Read →
Frames the failures inside the broader tech-industry accountability pattern, citing Meta's $18B youth addiction settlement as a precedent, and includes OpenAI's parental controls timing defense without editorial softening.
Parental alerts, mental health help, and study restrictions regularly drop the ball.
Originally reported by axios.com
Read the original article →Original headline: Common Sense Media Rates ChatGPT for Teens 'Unacceptable Risk,' Says Crisis Alerts Failed in 60-Min Tests