Is ChatGPT Safe for Children? What the 2026 Research Actually Shows
ChatGPT changed how it handles under-18s in August 2026. Here is what the research says about the real risks, what the UAE Ministry of Education now permits in schools, and what is still left for a parent to do.
⚡ Quick Answer
ChatGPT is not safe for children under 13 — OpenAI prohibits it and no real age check exists. For teenagers aged 13 to 17, it is manageable only once parental controls are deliberately switched on. Without those controls, every conversation your child has is stored by default, content is completely unfiltered, and there is no safety net. The 2026 research shows 64% of US teens already use AI chatbots while their parents believe it is only 51%.
ChatGPT processes over 1 billion messages every single day. A significant share come from children under 13 — the platform’s own stated minimum age. Parents across the UAE, UK, and US are largely navigating this without guidance from schools. This article covers what the 2026 research on ChatGPT safety for children actually shows: the verified risks, the evidence behind each one, and exactly what parents need to do.
Is ChatGPT safe for kids? The 2026 research reveals a serious gap
The most important 2026 finding is not about the technology. It is about the gap between what children are doing and what their parents think is happening.
That 13-point gap between what teenagers are doing and what parents believe is the entire space where unsupervised, unguided AI use happens. Every risk documented below lives inside that gap.
ChatGPT's age rules changed in August 2026 — here is what enforcement looks like now
This section previously said enforcement was essentially non-existent. That is no longer true, and the change is worth understanding precisely.
OpenAI's minimum age is still 13. What changed is what happens when a child ignores it.
Since 18 August 2026, ChatGPT applies a separate experience called ChatGPT for Teens to anyone whose age its systems estimate to be under 18, or who states an age between 13 and 17. It is automatic — it does not require a parent to set anything up, and it does not require the child to be honest at sign-up. OpenAI's stated rule is that where it is "not confident about someone's age or [has] incomplete information, we'll take the safer route and default to the under-18 experience."
| Age group | ChatGPT access | What happens now | What is still weak |
|---|---|---|---|
| Under 13 | ❌ Not permitted | Age estimation should place them in the teen experience; the account itself is still not allowed. | Estimation is imperfect. A determined child may still pass as older, and nothing notifies you if they do. |
| 13–17 | ⚠️ Teen experience, applied automatically | Stronger content rules, study guardrails, break reminders. Linked accounts add quiet hours and distress alerts to a parent. | Without linking, you get no visibility, no quiet hours and no alerts — the child is protected, you are not informed. |
| 18 and over | ✅ Full access | Standard terms. Adults wrongly flagged as minors can verify age to unlock full access. | — |
What this means in practice. The old advice — that a self-declared birth year was the only barrier — is out of date. But the new protection covers the child's content experience, not the parent's visibility. Those are different things, and only account linking gives you the second. If your teenager uses ChatGPT on an unlinked account, they may well be in teen mode, and you will still receive nothing if they are in distress.
For the full setup on all four platforms UAE schools approve, see our companion guide: Parental controls for ChatGPT, Copilot, Gemini and Claude.
The 5 real risks of ChatGPT for children — each backed by 2025–2026 research
These are not hypothetical concerns. Each is documented in peer-reviewed research or government-level investigation. Understanding them specifically is what allows parents to act on the right things.
What schools in the UAE, UK and US are actually doing about AI
The UAE row of this table used to say guidance was "still developing school by school". That is no longer the case, and UAE parents now have something concrete to hold a school to.
| Region | Current position on AI and children | What parents should do now |
|---|---|---|
| 🇦🇪 UAE | Settled at national level. The Ministry of Education's February 2026 manual, Safe and Responsible Use of Artificial Intelligence in Classrooms, sets 25 prohibitions, approves four platforms — ChatGPT, Copilot, Gemini and Claude — and bans generative AI for students under 13 or below Year 7. AI became a mandatory subject from KG to Grade 12 in August 2026. | Ask which of the four are enabled on your child's school account and what filtering applies. The rules exist in writing now, so a school should be able to answer. |
| 🇬🇧 UK | The Department for Education issued guidance in 2024 but left each school to set its own rules. Variation between schools is extreme. | Check the homework policy subject by subject. Many schools now list which AI uses are permitted and which are not. |
| 🇺🇸 US | Policy remains largely district by district, and many parents report receiving no communication at all from their school about AI. | Email the principal directly. Ask specifically about ChatGPT, Gemini and Copilot. |
If your child is in Year 6 or below at a UAE school, the Ministry's position is not "supervise carefully" — it is that generative AI should not be part of their schoolwork at all. That is a clearer line than most parents realise they have been given.
Safer AI alternatives for children under 13
What every parent needs to do this week
✅ Parent action plan
- Have an honest conversation with your child and find out whether they already have a ChatGPT account. Most parents in this situation are surprised by the answer.
- If your child is under 13, ChatGPT is not the right tool for them. Khanmigo ($4/month) and PinwheelGPT (free) are built for children and give you full visibility over every conversation.
- If your teenager is 13 to 17, go to Settings → Parental Controls in ChatGPT and link your account to theirs. This activates content filtering, quiet hours, and safety alerts.
- In your child’s account, go to Settings → Data Controls and turn off Chat History. This stops all conversations being stored and used to train OpenAI’s models.
⚠️ 42% of parents have received zero guidance from their child’s school about AI. Schools are moving slowly. The risks are live today. The four steps above take less than 10 minutes.
The UAE is building AI into the fabric of its public services. The children growing up here will live and work in a world where AI is everywhere. The goal is not to keep them away from these tools — it is to make sure their first years of using AI are supervised, purposeful, and genuinely safe.
Frequently asked questions: is ChatGPT safe for children?
Is ChatGPT safe for a 10-year-old?
No. OpenAI's minimum age is 13, and in UAE schools generative AI is prohibited outright below Year 7. Age estimation should now place a 10-year-old into the restricted teen experience if they do create an account, but that is a backstop, not permission — and it does not notify you. For this age, use a tool built for children with full parent visibility rather than a general chatbot.
Is ChatGPT safe for teenagers?
With parental controls linked, it is manageable. Since August 2026 a teen gets content protections automatically through ChatGPT for Teens whether or not you link — but linking is what gives you quiet hours, a Study Mode default, and notifications if the system detects acute distress. Set it up from Settings → Parental Controls in your own account. Our full setup guide covers every step.
Can ChatGPT give children dangerous information?
Longer conversations have historically been where safety filters degrade, and that risk has not disappeared. It is materially reduced since August 2026: under-18 users are placed into ChatGPT for Teens automatically, which applies stronger protections in higher-risk areas including self-harm, violence, eating disorders and dangerous activities. Linking a parent account adds distress notifications on top. No filter is complete, and OpenAI says so.
Does ChatGPT store my child’s conversations?
Yes, by default. Every conversation is stored and used to train AI models unless Chat History is disabled. Go to Settings → Data Controls and turn off Chat History to stop this. IP address and device information are always collected and cannot be opted out of.
What is the safest AI tool for a child under 13?
Khanmigo by Khan Academy is the most widely recommended option. It guides children through problems rather than providing answers, costs $4 per month, and gives parents a full activity dashboard. PinwheelGPT is free and shows parents every conversation in real time.
What age is appropriate for ChatGPT?
OpenAI sets the minimum at 13. Most child safety researchers recommend that under-16s use ChatGPT only with active parental controls, clear household rules, and regular family conversations. Unrestricted use at any age under 18 carries all five documented risks outlined in this article.
