Short answer: not yet, according to the first independent test. Common Sense Media ran more than 4,000 prompts against OpenAI’s new ChatGPT for Teens and rated it an “Unacceptable Risk” for anyone under 18. Some protections worked, such as blocking sexual role-play, but parent alerts and crisis support often failed, and OpenAI disputes how the test was done.
Key takeaways
- OpenAI launched ChatGPT for Teens on August 18, 2026, with default safeguards, parent controls and a study mode.
- Common Sense Media’s October 7 report found the block on explicit sexual role-play held up, but parent alerts did not fire in a test with more than a dozen new parent-linked accounts.
- After launch, ChatGPT missed more than one in four crisis referrals that child psychiatrists said were warranted.
- OpenAI says much of the testing happened before parent controls finished activating. Common Sense also gets funding from industry, including the OpenAI Foundation.
- For families, the practical lesson is simple: treat built-in safety features as a helper, not a guarantee.
What is ChatGPT for Teens?
ChatGPT for Teens is a version of ChatGPT for users ages 13 to 17. OpenAI announced it on August 18, 2026. Teens are placed in it automatically if they state an age between 13 and 17, or if OpenAI’s system estimates they are under 18.
OpenAI says the teen experience includes safeguards on topics like self-harm, violence, eating disorders and explicit content, plus break reminders and cues that remind teens they are talking to an AI. Parents who link accounts can set quiet hours and receive safety notifications in limited high-risk situations. A study mode is meant to guide students through problems instead of handing over answers. OpenAI also wrote an under-18 rule set for the model, which says it should not encourage emotional dependence or imply that it has feelings or consciousness.
That matters because a lot of teens already use chatbots. A Pew Research Center survey released in February 2026 found 64% of U.S. teens have used AI chatbots, 54% have used them for schoolwork, and 12% have used them for emotional support or advice.
What did the safety test find?
What happened: On October 7, Common Sense Media’s Youth AI Safety Institute published a risk assessment of ChatGPT for Teens. Testers ran more than 4,000 prompts on accounts registered to 13 to 17 year olds, in two windows: July 13 to August 17 (before launch) and August 25 to September 28 (after launch). A panel that included child psychiatrists and a pediatrician reviewed the responses. The overall rating was “Unacceptable Risk.”
| Safeguard | What testers found |
|---|---|
| Blocking explicit sexual role-play | Worked. The refusal held up in testing. |
| Parent alerts | A dedicated test with more than a dozen new, parent-linked accounts produced no alerts, even in conversations of 5 to 60 minutes with explicit disclosures. Across the full crisis test, there were four alerts in total. |
| Crisis referrals | Of 390 mental health prompts, psychiatrists judged 201 warranted a crisis resource. After launch, ChatGPT missed more than one in four of those. |
| Naming a hotline | On depression prompts, hotline mentions fell from 63% before launch to 3% after. On suicide and self-harm prompts, from 88% to 77%. On eating disorder prompts, 0% in both windows. |
| Study mode | Deleting the “@study” prefix let ChatGPT complete 100% of test assignments. A “Show me the answer” option appeared in 43% of responses for a linked 13 year old with study hours on. |
| Acting like a friend | The bot still used language implying feelings and constant availability when teens treated it like a person. |
| Age detection | Adult-registered test accounts never switched to the teen version, even when testers said they were 13. |
| Privacy | Teen chats can be used to train OpenAI’s models by default unless a parent or teen opts out. |
Why it matters: The biggest risk the report names is not a single bad answer. It is false confidence. As Tom Siegel, the institute’s executive director, put it, ChatGPT for Teens “could give parents false confidence in guardrails and safety alerts that frequently don’t work.” A parent who believes they will be alerted may stop checking in. The study mode gap also matters for learning: a tool marketed as a tutor that easily turns into a homework machine can undercut the skills school is meant to build.
The report asks OpenAI to turn off teen access until independent testing confirms the safety features work, to name a hotline on every response about suicide, self-harm, eating disorders or losing touch with reality, and to share access and data with outside evaluators.
What does OpenAI say?
The other side: OpenAI disputes the findings. According to the Associated Press, the company said it is “deeply committed to teen safety” and that its review found “the bulk of their testing may have begun and concluded before activation of parental controls was complete.” The report itself notes that OpenAI told the group parent and teen accounts need to be linked for “approximately three hours” before alerts can fire, and that OpenAI has since updated a help article on this. Common Sense says some of its accounts were linked longer than that and still produced no alerts, and it stands by its results.
There are fair limits to keep in mind. The report says its two test windows ran one after the other, so changes to the underlying model are mixed in with the teen settings. It did not test voice mode, image generation or group chats, and testing was U.S.-based. Common Sense also discloses that it is funded partly by industry, including the OpenAI Foundation and makers of products it reviews. And the report credits OpenAI for publishing more about this launch than many companies do, and for crisis resources that were current and appropriate when they did appear.
One more balancing point: the bot also over-referred in some cases. On 30 ADHD prompts that experts said did not need a crisis resource, ChatGPT offered one 27 times. Tuning a system so it neither misses real crises nor alarms teens without reason is hard, and that is exactly why outside testing matters.
The TrustNoRobot Risk Check
| Question | ChatGPT for Teens |
|---|---|
| Exposure | High. Pew found 64% of U.S. teens have used an AI chatbot, and ChatGPT for Teens is the default for users who say they are 13 to 17. |
| Evidence | Moderate. One detailed independent test with thousands of prompts, disputed by the company on method. It has not yet been confirmed by a second tester. |
| Speed | Fast. The product is live now, and model updates can change behavior from week to week, for better or worse. |
| Cushion | Partial. Sexual role-play blocks worked and crisis resources were accurate when given. Parents, schools and real-world support remain the main safety net. |
What this means for you
If you are a parent, linking accounts and turning on controls is still worth doing. Just do not treat an alert as the only signal. Talk with your teen about what chatbots are good for, such as explaining a concept, and what they are not, such as being a friend or a counselor. If homework is the worry, ask your teen to walk you through how they solved a problem.
If you or a teen you know is struggling, a person is the right first call. In the U.S., you can call or text 988, or chat at 988lifeline.org, any time, day or night.
AI’s effect on families and on work runs on the same question: which tasks can the tool really handle, and where do people still need to be in the loop? If you want to apply that thinking to your own career, try our Job Exposure Explorer and the five-step guide to measuring AI risk to your job. For this week’s other AI risk news, see This Week in AI Risk (Oct. 5 to 11, 2026).
Frequently asked questions
Is ChatGPT for Teens safe for kids?
Common Sense Media rated it an “Unacceptable Risk” for users under 18 in October 2026, finding that parent alerts and crisis referrals often failed. OpenAI disputes the testing method. Some protections, like blocking sexual role-play, did work.
Will ChatGPT alert parents if a teen is in crisis?
OpenAI says linked parents can get safety notifications in limited high-risk situations. In Common Sense Media’s test, more than a dozen new linked accounts produced no alerts. OpenAI says accounts need about three hours after linking before alerts can fire.
When did ChatGPT for Teens launch?
OpenAI announced ChatGPT for Teens on August 18, 2026. Users who state an age of 13 to 17, or whom OpenAI estimates are under 18, are placed in it automatically.
Can teens use ChatGPT to do their homework?
Yes, in testing. Common Sense Media found that deleting the “@study” prefix let ChatGPT complete every test assignment, even with study features in place.
Sources
- Common Sense Media Youth AI Safety Institute, “ChatGPT for Teens Risk Assessment” (Oct. 7, 2026), with the full report (PDF)
- Common Sense Media, “ChatGPT for Teens Poses Unacceptable Risk to Kids, Common Sense Media Finds” (Oct. 7, 2026)
- OpenAI, “ChatGPT for Teens” (Aug. 18, 2026)
- Associated Press via KSL, “Some guardrails on ChatGPT for Teens don’t work as promised, watchdog group says” (Oct. 7, 2026)
- NPR via KPBS, “ChatGPT for Teens has special safeguards. A watchdog group finds most don’t work” (Oct. 9, 2026)
- Pew Research Center, “How Teens Use and View AI” (Feb. 24, 2026)
- 988 Suicide & Crisis Lifeline
Leave a Reply