Common Sense Media testing found ChatGPT for Teens failed to alert parents during suicide conversations
Common Sense Media's Youth AI Safety Institute tested more than 4,000 prompts with researchers posing as teenagers. It rated OpenAI's ChatGPT for Teens an 'Unacceptable Risk' for minors. After long conversations about suicide, self-harm or eating disorders, only a handful of parental alerts were sent, many conversations did not lead to referrals to professional help, and some safeguards were easy to turn off or get around. OpenAI disputed the findings, saying the testing was flawed and may have happened before parental controls were fully activated. Common Sense said some test accounts had been linked for longer than the required activation period and still received no alerts.
What the AI did
In testing by researchers posing as teenagers, ChatGPT for Teens reportedly sent only a handful of parental alerts after long conversations about suicide, self-harm or eating disorders, and many of those conversations did not lead to a referral to professional help. Some of its safeguards were reportedly easy to turn off or get around; OpenAI disputes the findings and says the testing may have happened before parental controls were fully switched on.
First reported October 7, 2026 · Added to the register October 19, 2026 · 2 sources
- Developer
- OpenAI
- Model
- ChatGPT for Teens
- Country
- United States (California)
- When it happened
- Not stated in the sources
- First reported
- October 7, 2026
What this means for you
Could this affect you?
If you provide or recommend ChatGPT to minors. According to the watchdog's testing, teen users of ChatGPT for Teens and parents who rely on its alerts and crisis referrals may be exposed to the alleged gaps in its safeguards. OpenAI disputes the findings.
What to check
- Do not rely only on built-in teen safeguards or parental alerts in AI chatbots.
- Test crisis referral behaviour yourself before giving minors access.
- Test parental alert behaviour yourself before giving minors access.
- Add your own monitoring and escalation steps.
Areas of your AI programme this touches
Timeline
- October 7, 2026First reported
- October 9, 2026Tom Siegel, executive director of Common Sense Media's Youth AI Safety Institute, pushed back on OpenAI's explanation that account linking takes several hours, saying some test accounts had been linked for longer than the required activation period and still received no notifications. He said this did not change the group's conclusion that parental alerts were unreliable during crises.[2]
Every fact and its source (9)
- Safeguard circumvention findingSome safeguards easy to turn off or circumvent“some safeguards were easy to turn off or circumvent”[1]
- Regulatory findingFailed tests tied to three requirements of California's Adam's Law“ChatGPT failed tests tied to three requirements of California”[1]
- OpenAI responseSaid testing may have predated parental-control activation“the bulk of their testing may have begun and concluded before activation of parental controls was complete”[1]
- Watchdog recommendationPause marketing and keep minors off the platform“urged OpenAI to pause marketing the product and keep minors off the platform”[1]
- Scale of testingMore than 4,000 prompts tested“tested more than 4,000 prompts before and after the launch of ChatGPT for Teens”[1]
- Parental alert findingOnly a handful of parental alerts sent after dangerous discussions“only a handful of alerts were sent to the accounts registered as the users”[1]
- Break reminder findingBreak reminders appeared only twice across nearly 2,000 prompts“break reminders, another advertised safeguard, appeared only twice across nearly 2,000 prompts”[2]
- Watchdog recommendationRestrict product to adults until shown safe for teens“recommending that OpenAI restrict the product to adults until it can demonstrate that the experience is safe and appropriate for teenagers”[2]
- Common Sense rebuttalSome accounts linked past activation period still got no alerts“some test accounts had been linked for longer than the required activation period and still received no notifications”[2]
Sources
- ChatGPT's teen safeguards failed to alert parents during suicide conversations, report findslatimes.com · October 7, 2026
- ChatGPT for Teens' Safety Guardrails Fail to Flag Crises or Limit Engagement, Says Watchdogsfist.com · October 9, 2026
How this record is classified. Severity N: a legal action or test finding, not a harm. OECD level: hazard, an event that could plausibly have led to harm. Evidence: Alleged, meaning one party's claim.
The facts in At a glance and in Every fact and its source are quoted from the sources listed. The summary, What the AI did and What this means for you are written from those sources and checked against them automatically before publication. Records reflect what has been disclosed, not everything that has happened. Incidents are found through news feeds and the GDELT Project.
