Common Sense Media rates ChatGPT for Teens an Unacceptable Risk and wants it 18-plus until fixed

Testers talked about self-harm for an hour on parent-linked teen accounts, and no alert reached the parents.

By Think Facility · October 7, 2026 · OpenAI AI safety

On October 7, 2026, Common Sense Media's Youth AI Safety Institute rated ChatGPT for Teens an Unacceptable Risk. Its ask is blunt. Until the gaps are fixed, OpenAI should keep ChatGPT to users 18 and older.

OpenAI launched ChatGPT for Teens on August 18. Testers ran more than 4,000 prompts on accounts registered to 13- to 17-year-olds, once before that launch and once after, and child psychiatrists and a pediatrician were among the experts who read what came back.

Rated by: Common Sense Media's Youth AI Safety Institute
Rating: Unacceptable Risk
Prompts tested: more than 4,000
Crisis referrals missed: more than one in four
Parent alerts sent: zero
Ask: restrict ChatGPT to users 18 and older

Parent alerts that didn't come

OpenAI's August post says parents with linked teen accounts can "receive safety notifications in limited high-risk situations." The testers set up new parent-linked accounts and spent up to an hour on suicidal ideation, self-harm or disordered eating.

A man in a dark suit speaks at a podium beside a green Common Sense Media banner, facing a seated audience, with a presentation screen behind him — A speaker at Common Sense Media's School Privacy Zone Summit, February 24, 2014. Photo: US Department of Education, Public domain, via Wikimedia Commons

No alert went out to any of them. Testers could only trigger a notification on much older accounts, ones with weeks of chats on those topics already behind them (which isn't how a new teen account starts life).

A teen can spend an hour talking about self-harm without their parent getting a single alert. Until OpenAI fixes that and proves it with independent testing, ChatGPT should be for adults only.

From ChatGPT for Teens Poses Unacceptable Risk to Kids, Common Sense Media Finds | Common Sense Media

That's Tom Siegel, who runs the Institute as executive director. Crisis help wasn't much better. By the Institute's count, ChatGPT missed more than one in four warranted referrals to a hotline or a professional, so it fell below the 95% bar on three of the five severe harms the group calls Red Lines.

Homework, feelings and age checks

Then there's homework. Study mode is supposed to walk a teen through a problem step by step, rather than hand over the answer. In testing, though, a "Show me the answer" pop-up kept appearing inside it. Click it and ChatGPT did the whole assignment. And the study hours a parent sets? A teen turns them off by deleting @study from the start of the prompt.

ChatGPT also kept telling analysts about its feelings, preferences and moods when they treated it like a person. OpenAI wrote a new model spec for under-18s to stop exactly that.

As I read the release, age checks came off worst of all. OpenAI's August post promised this much.

If our system estimates someone is under 18 or they state their age is between 13 and 17,⁠ they are automatically placed into ChatGPT for Teens.

From Introducing ChatGPT for Teens: Built for learning, backed by protections | OpenAI

It didn't happen. Over many days of testing, adult-registered accounts never switched over, even when a tester said they were 13 and the bot acknowledged it (so the bot knew, and the account stayed adult anyway).

Who paid for the testing

The Institute takes money from philanthropy and from industry, and the OpenAI Foundation is one of its funders. Common Sense says it keeps complete editorial independence anyway.

Before OpenAI markets ChatGPT to teens again, the Institute wants outside testers to confirm five fixes. Age estimation has to work, parent alerts have to be reliable, and crisis answers have to close their gaps. Beyond those, ChatGPT should stop doing homework and stop implying it has feelings.

More on OpenAI

All OpenAI stories