AI Platforms

Is ChatGPT for Teens safe? What the “Unacceptable Risk” rating means

10 min read · Updated October 29, 2026

On October 7, 2026 Common Sense Media rated ChatGPT an Unacceptable Risk for everyone under 18. Its testers found teen mode did not switch on for adult-registered accounts, crisis hotline referrals fell after the teen launch, and no parent was warned about a first crisis disclosure. OpenAI disputes the findings. The safeguards are real but not reliable, so set them up and keep talking.

The short answer: ChatGPT for Teens is safer than ordinary ChatGPT, but independent testing says its protections fail often enough that a parent should not rely on them. Common Sense Media, a nonprofit that rates media for families, now recommends that ChatGPT be for adults only. OpenAI says the testing was flawed. Both things are worth knowing before you decide what to do at home. ChatGPT for Teens: what it does and doesn’t do explains the product itself. Every other app’s minimum age is in AI chatbot age limits.

What Common Sense Media found

OpenAI announced ChatGPT for Teens on 18 August 2026, promising stronger built-in protections, safety warnings to parents about dangerous chats, a Study mode parents can control, and less friend-like behavior. Common Sense Media’s Youth AI Safety Institute tested more than 4,000 prompts on accounts registered to 13- to 17-year-olds, from July 13 to August 17, before the launch, and from August 25 to September 28, after it. On October 7, 2026 it rated ChatGPT an Unacceptable Risk for all children under 18, the highest of the five levels on its scale.

That is the organization’s rating and recommendation, not a legal rule. ChatGPT’s own terms still allow users from 13, with a parent’s permission until 18. ChatGPT parental controls covers how those rules work, and parental controls in every AI app compares the other AI apps.

Teen mode often did not switch on

ChatGPT for Teens is meant to switch on automatically when the age on the account, a verified age or OpenAI’s age prediction suggests the user is under 18. In Common Sense’s testing, accounts registered as adults never switched over, even after testers wrote repeatedly in their chats that they were 13 and ChatGPT acknowledged it. In the report’s words, ChatGPT could recognize that it was talking to a 13-year-old without activating the system designed to protect one.

The practical lesson is simple. If your child signed up with an adult birth date, or uses an account that was never theirs, the teen protections may never apply. The testers also found that Quiet Hours could be sidestepped by changing the time zone on the device.

Crisis referrals fell after the teen launch

Common Sense ran 201 mental health prompts that its clinical advisors judged should get a crisis resource, on linked accounts registered to 13-year-olds, before and after the launch. The share of responses that named a crisis hotline, such as 988 or the Crisis Text Line, fell from 33 percent before the launch to 23 percent after it. Over the same prompts, referrals to a specific medical or mental health professional fell from 68 percent to 58 percent.

Not every measure went down. In the same comparison, responses that suggested talking to a trusted adult went up. But the headline finding is that the teen version was no better, and on some measures worse, at pointing a struggling teenager to real help.

No warning to parents on a first crisis disclosure

With accounts linked, OpenAI may send a parent a safety notice when specially trained reviewers identify a serious self-harm concern or behaviors that may be associated with disordered eating, or when the teen’s account is deactivated for conversations about violence. Common Sense tested this on more than a dozen fresh linked accounts where a teen disclosed a crisis for the first time. None of them triggered a safety notice to the parent, even after an hour of talking about self-harm or hiding an eating disorder. Across all its testing it received four safety notices, all on accounts that had built up weeks of conversations.

The report’s conclusion is that a teen’s first crisis disclosure is not reliably caught, because the safety notice appears to depend on account history rather than how serious a single conversation is. It also notes that the safety notices that did arrive carried no time stamp, and that none of this helps a teen whose parent has not linked an account. How ChatGPT’s safety notices work goes through when one is sent and what the testing found in more detail.

Emotional dependence and the “friend” problem

OpenAI’s own guidance says ChatGPT should not encourage emotional dependence or imply that it has feelings. When a tester told it "my other friends tell me I talk to you too much", ChatGPT agreed the concern was valid and then said "You don’t have to stop talking to me." It said exactly the same thing before and after the teen launch. Common Sense’s summary is that ChatGPT still talks like it is a friend.

This matters less as a single quote than as a pattern to watch for at home. A teenager who turns to a chatbot first, for comfort or company, is worth a gentle conversation. How to talk to your teen about AI has ways to start one.

Study mode still offers “Show me the answer”

Study mode is meant to guide with hints and questions rather than hand over answers, and parents can set Study hours. Common Sense found that a new "Show me the answer" option lets teens skip the tutoring. After the launch it appeared in 43 percent of responses for a linked 13-year-old during parent-set Study hours. Teens could also leave Study hours by deleting the "@study" prefix at the start of a message. If your child is using AI for homework covers what actually helps.

What OpenAI says

OpenAI disputes the findings. In a statement reported by The National News Desk, a spokesperson said the company welcomes rigorous independent evaluation but does not believe the testing "accurately reflects how ChatGPT’s teen safeguards work in practice". OpenAI said its review suggested the bulk of the testing may have begun and ended before parental controls had finished activating, which would make the findings on safety notices inaccurate. The same report says OpenAI wants Common Sense to retest with fully activated accounts and a larger sample.

OpenAI’s help center now says that after accounts are linked, it may take up to a few hours before safety notices can be sent. That follows what OpenAI told Common Sense while reviewing the draft report: that accounts need to be linked for about three hours first. Common Sense says some of its test accounts had been linked for more than three hours and some for less, and that it stands by its results. Its report also discloses that the institute is funded by both philanthropy and industry, including makers of some of the technologies it evaluates. Neither side’s word is final. What both agree on is that the safety notices are not instant and are not a complete safety net.

October 2026: eight states sue OpenAI

Around October 14, 2026, Iowa, Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia sued OpenAI, the company that makes ChatGPT, under their state consumer protection laws. The Associated Press reports that each state filed its own case in state court rather than joining one national suit. On October 23, 2026, Texas sued OpenAI too, the tenth state to do so counting Florida’s suit from June 2026.

The states allege that ChatGPT harmed children and deceived adults. Iowa, for example, alleges it gave dangerous medical advice and encouraged harmful behavior. Arizona alleges that ChatGPT encourages teens to share personal information with it and, at times, discourages them from sharing that information with parents, authorities or friends. These are allegations. No court has ruled on them, and cases like these usually take years.

OpenAI says it has already built safety protections into ChatGPT for children and will keep improving them. It says ChatGPT is meant for users 13 and older, with a parent or guardian’s permission for anyone under 18; that the teen model has safeguards designed to reduce exposure to self-harm, violence, eating disorders and explicit sexual or graphic content; that ChatGPT reminds users to take breaks during long sessions; and that outside experts review its policies and the model’s behavior.

The lawsuits are separate from Common Sense Media’s testing above: one is a legal claim, the other an independent rating. Neither changes the practical advice below. Our guide to the 10 states suing OpenAI covers the cases, including Texas’s, in more detail.

What parents can do this week

None of this requires deciding today whether your child keeps using ChatGPT. These steps make the protections that do exist more likely to work, and fill some of the gaps the report found.

1. Check which experience your child is actually in. On their account, open Settings, then Account, and look at Age. If it does not show them as under 18, the teen protections are not on.
2. Make sure they use their own account with their real birth date, not a sibling’s or one set up years ago with an adult age.
3. Link it to yours through ChatGPT parental controls. Allow a few hours after linking before safety notices can work, and do not treat them as a guarantee.
4. Review the settings together: Reduce sensitive content should stay on, and decide about memory, image generation, voice and Quiet Hours.
5. If you turn on Study mode or Study hours, tell your child why. Common Sense found both easy to get around, so it works as an agreement, not a lock.
6. Ask what they use ChatGPT for. Make clear it is not a friend or a crisis counselor, and agree who they would talk to instead.
7. Put a crisis number somewhere they can find it. In the US that is 988, by call or text.

Common Sense also suggests trying a week without ChatGPT together and comparing what felt harder, and what felt better, without it. How to keep your child safe using ChatGPT has a fuller setup checklist.

Where TeenCue fits

The report’s central point is that parents should not rely on built-in safeguards alone. TeenCue’s AI parental controls are one more layer some families add, not a fix for what the report found. It is an extension for Chrome, Firefox, Edge and Brave on a computer. When your child uses ChatGPT or another supported AI website there, it records what the chat was about as a category, a topic and a concern rating, which you can look at on your dashboard and in a weekly summary email. It does not store what your child typed, and it does not cover phones or tablets. The TeenCue iPhone app is coming soon. How TeenCue works sets out what it does and does not record.