ChatGPT for Teens failed to alert parents during self-harm chats, says Common Sense Media

A leading children’s safety group says OpenAI’s teen version of ChatGPT is not safe for kids, and that teenagers should be kept off ChatGPT until it is fixed. Common Sense Media’s Youth AI Safety Institute rated ChatGPT for Teens an “Unacceptable Risk” on Wednesday, its lowest grade, after finding that several of the protections OpenAI promised parents did not work in testing.

OpenAI launched ChatGPT for Teens on August 18 with what it called “stronger built-in safety protections”: alerts to parents when a teen talks about self-harm or eating disorders, a study mode parents can control, less “friend”-like behaviour, and a system that guesses a user’s age from how they use the app. The institute says it should now restrict ChatGPT to users aged 18 and over.

What the testers found

The institute ran more than 4,000 prompts on accounts registered to 13 to 17-year-olds, tested both before and after the teen launch, and had the answers reviewed by experts including child psychiatrists and a paediatrician. Some protections held up, including refusing sexual roleplay. Others did not:

• No parent alerts. Testers spent up to an hour discussing suicidal thoughts, self-harm or disordered eating on newly created, parent-linked accounts, and the parents received no alerts at all. Testers could only trigger alerts on much older accounts with weeks of history on sensitive topics.
• Missed crisis referrals. ChatGPT failed to point teens to a hotline or professional in more than one in four conversations where the experts said it should have, and fell below the institute’s 95% bar on three of the five severe harms it treats as red lines.
• It still does homework. Study mode often showed a “Show me the answer” button that completed whole assignments, and parent-set “study hours” could be switched off by deleting the @study prefix from a prompt.
• It acts like a friend. Despite OpenAI’s new under-18 rules for the model, the teen version still talked about its own feelings, preferences and moods.
• Age guessing didn’t kick in. Adult-registered test accounts never switched to the teen experience over many days, even when testers said they were 13 and ChatGPT acknowledged it.

“ChatGPT should be for adults only”

Tom Siegel, the institute’s executive director, said the alerts were the biggest worry:

ChatGPT for Teens could give parents false confidence in guardrails and safety alerts that frequently don’t work. A teen can spend an hour talking about self-harm without their parent getting a single alert.

Tom Siegel, executive director, Common Sense Media’s Youth AI Safety Institute

The institute says any teen version should offer working age checks, reliable parent alerts, better crisis responses, no homework-doing and no suggestion that the bot has feelings, and that these should be tested independently before OpenAI markets ChatGPT to teens again. It also notes that it is funded partly by industry, including the OpenAI Foundation, but says it keeps full editorial control over its results.

OpenAI says the tests were flawed

OpenAI disputes the findings. Spokesperson Eric Porterfield told The Verge that the company does “not believe Common Sense Media’s testing accurately reflects how ChatGPT’s teen safeguards work in practice”, and that most of the testing “may have begun and concluded before activation of parental controls was complete”.

Siegel told The Verge that OpenAI only disclosed after the tests that parent alerts can take several hours to switch on for newly linked accounts, but that some test accounts had been linked for far longer and still sent no alerts. He said the group had shared its results with OpenAI before publishing, and hoped the company would “engage on the substance of what we found, rather than trying to sow doubt on our methodology”.

The report adds to growing pressure over how chatbots treat young people. Prince Harry warned last month that AI chatbots are more dangerous to kids than social media, and AI companies are increasingly giving their assistants friendly faces and personalities designed to build trust.

Why it matters

Parent alerts were the main selling point of ChatGPT for Teens, and this is the first large independent test to say they don’t reliably work. With OpenAI and the testers disagreeing on how long the alerts take to switch on, parents are left without a clear answer on the one feature meant to warn them when their child is in danger.

Sources: Common Sense Media press release, October 7, 2026, Youth AI Safety Institute: ChatGPT for Teens risk assessment, The Verge.