Kids and AI in Singapore: Ages, Controls and School Rules

By the WahLiao desk · Last verified 3 October 2026

Here’s the short answer. Most Super Intelligence chatbots, the kind of AI behind ChatGPT, Claude and Gemini, aren’t meant for young children to use on their own. ChatGPT, Microsoft Copilot and Meta AI set 13 as the minimum age, and ChatGPT asks under-18s to have a parent’s permission. Claude is for those 18 and over. Gemini is the exception: a parent can switch it on for a child under 13 through Google Family Link. In Singapore schools, children in Primary 1 to 3 don’t use chatbots for schoolwork; supervised, MOE-built tools start from Primary 4, inside the Student Learning Space (SLS).

If you remember one thing, make it this: if your child uses a chatbot at home, link it to your own account with the parental controls the provider offers, and keep the device in a shared room rather than a bedroom. Controls help, but you being nearby helps more.

Children and AI in Singapore: quick facts

ChatGPT	13+, parent’s permission under 18; parental controls since 2025
Gemini	Under-13s only if a parent turns it on in Family Link; 13 to 17 get teen protections
Claude	18+ only
Copilot and Meta AI	13+; parents can limit Copilot through Microsoft Family Safety
In school	No chatbots P1 to P3; supervised SLS tools from P4; more independent use in secondary
Screen time (MOH)	None under 18 months; under 1 hour a day for ages 3 to 6; under 2 hours for 7 to 12, outside schoolwork
If images are misused	Report to the Online Safety Commission; parents can file for under-18s

The WahLiao Verdict

Under 13	Only with you beside them, and only where the tool allows it.
Teens	Link accounts, set quiet hours, and chat about what it gets wrong.
Homework	A tutor, not a ghostwriter. If the teacher says no AI, it’s no AI.
Companion bots	Best left out of childhood. A chatbot isn’t a friend.
Photos of your child	Keep them out of chatbots and image apps.

Minimum ages and parental controls, tool by tool

Each provider sets its own age rule in its terms, and they differ more than most of us would guess. None of them checks age very hard at sign-up, so in practice the rule is yours to keep.

Tool	Minimum age	What parents can control
ChatGPT	13; under 18 needs a parent’s or guardian’s permission	Link accounts; quiet hours; switch off memory, voice mode and image generation; sensitive-content reduction on by default; safety alerts in rare serious cases
Gemini	Under 13 via Family Link, if a parent turns it on; 13 to 17 get “Gemini with teen protections”	Switch Gemini on or off for each child in Family Link; parents get an email when the child first uses it
Claude	18 (or older if local law requires)	Not designed for children at all
Microsoft Copilot	13	Block the app, limit time and switch off features such as image generation, voice and AI companions through Microsoft Family Safety or device controls
Meta AI	13	Teen protections by default; access to AI characters for teens paused since January 2026

Two details are worth knowing. ChatGPT’s parental controls don’t let you read your teen’s conversations; they shape the settings, and OpenAI may notify you only in limited situations where there’s a serious safety concern. And Google says its filters for children are “not perfect”, which, as far as we can tell, is a fair description of all of them. Independent testers such as Common Sense Media have rated Gemini’s teen version as high risk overall, with unsafe mental-health and emotional support as their most serious concern. Our explainer on how Super Intelligence works covers why these systems can sound sure and still be wrong.

How Singapore schools use Super Intelligence

MOE’s approach is age-appropriate and unhurried. At lower primary, learning is “print-first”, with hands-on work and little technology; MOE says P1 to P3 pupils don’t use AI directly, and schools don’t require them to use chatbots for schoolwork. From Primary 4, AI use is “structured and limited, in class and under close teacher supervision”, using MOE-developed tools rather than general-purpose chatbots. In secondary school, students use their personal learning devices and AI more independently, through structured tasks and project work.

The tools sit inside SLS. MOE names the Learning Assistant (LEA), feedback assistants such as Short Answer Feedback, and an Adaptive Learning System. The Learning Assistant comes with guard rails: a teacher has to switch it on and can cap the number of interactions, and it nudges students to think things through rather than simply handing over answers.

On homework and tests, MOE has been consistent through 2026. Where a task assesses what a student can do alone, “students will need to complete work without AI assistance”, and proctored examinations disallow AI. Teachers may allow it for some activities, such as generating ideas before a class discussion. So the rule at home is a simple one: follow the teacher’s instruction for that piece of work. MOE says it shares resources and tips for parents through Parents Gateway, and some schools send home “A Parent’s Guide to Generative AI tools for Learning”.

The rules protecting children online

There are three layers worth knowing about. First, IMDA’s Code of Practice for Online Safety for App Distribution Services covers the Apple App Store, Google Play, Huawei AppGallery, Microsoft Store and Samsung Galaxy Store. It took effect on 31 March 2025, and the stores had to have age assurance in place by 31 March 2026, at minimum working out whether a user is under 18 so that age-inappropriate apps can be restricted. MDDI has said such measures are “useful but insufficient” on their own.

Second, the Online Safety Commission began operating on 29 June 2026 under the Online Safety (Relief and Accountability) Act 2025. For now it handles five harms: online harassment, doxxing, online stalking, intimate image abuse and image-based child abuse, with eight more categories to follow. For intimate image abuse, image-based child abuse and doxxing, victims can report directly at osc.gov.sg; for harassment and stalking, they need to report to the platform first. Parents and guardians can file on behalf of anyone under 18. The Commission can order content removed or accounts restricted. Malay Mail reported that it received more than 500 reports in its first two months.

Third, MOH’s Grow Well SG guidance on screens: no screen use for infants up to 18 months, screens only for learning from 18 months to six years (under one hour a day outside school for ages three to six), and under two hours a day for ages seven to 12 unless it’s schoolwork. Chatbot time counts as screen time too.

These rules can change, so check the official page before you act. If we’ve got something wrong, tell us and we’ll fix it.

Practical rules for parents

Homework

Good uses: asking the chatbot to explain a concept a different way, quizzing your child before a test, or checking an answer after they’ve had a go. Less helpful uses: asking it to write the composition or solve the problem sum outright. One habit we like is “try first, ask second, explain back”: your child attempts the work, asks the tool only when stuck, then explains the answer to you in their own words.

Companion chatbots

Apps that play a friend, partner or character are built to keep people chatting, and they’re a poor fit for children. Meta paused teen access to its AI characters in January 2026, and Microsoft lets parents switch off Copilot’s AI companions. Keep children on general assistants with parental controls, and talk openly and kindly if they start treating a chatbot as a confidant. If your child is struggling, a trusted adult, teacher or school counsellor is the right first stop, not a bot.

Images of children

Please don’t upload your child’s photos to chatbots, “AI photo” apps or face-swap tools, and think twice before posting school-uniform photos publicly. What you type and upload is stored by the provider; our guide to what happens to what you type walks through the settings.

Deepfakes

Explain to older children, calmly, that fake pictures, voices and videos of real people are easy to make, and that creating or sharing a sexualised fake of anyone, including a classmate, is serious. If your child is targeted, keep a record, don’t share the material further, and report it to the platform and the Online Safety Commission. See deepfakes and the law and how to spot a voice-clone scam for the family version of that talk.

Children and AI: FAQ

What age can my child use ChatGPT in Singapore?

OpenAI’s terms set 13 as the minimum, and anyone under 18 needs a parent’s or guardian’s permission. You can link accounts and set quiet hours and other limits.

Can my child under 13 use Gemini?

Only if you turn it on in Google Family Link for their supervised account. Google says rollout is gradual and its child filters aren’t perfect, so it’s best to stay close by.

Can I read my teenager’s ChatGPT chats with parental controls?

No. The controls change settings such as quiet hours, memory, voice and image generation. They don’t show conversation history, though OpenAI may alert you in rare serious safety situations.

Are Singapore students allowed to use AI for homework?

It depends on the task. MOE says work that assesses independent mastery must be done without AI, and exams disallow it. Teachers may allow AI for some activities, so follow the instruction for each assignment.

What should I do if a fake or intimate image of my child is shared online?

Don’t forward it. Report it to the platform and to the Online Safety Commission at osc.gov.sg, which accepts reports from parents on behalf of children under 18 and can order removal.