Learn Coding

8 min

Safe AI tools for kids: how to choose

Published: 15.09.2026·Updated: 15.09.2026

Fajar Hidayat

Coding Curriculum Team, Algonova

Safe AI tools for kids share five traits, and you can judge any app against them in about two minutes without installing anything. The list below covers ages 6 to 17, names what to skip, and includes the settings to change on day one. One rule matters most: if the tool needs an account and your child is under 13, it is not for them.

The five point check

Run any AI app through these five points before your child touches it.

Check	What to look for
Age limit	A stated minimum age you can find in two clicks
Account	Can it be used without a personal login for your child?
Data	Does it need photos, voice or personal details to work?
Open chat	Can it produce anything at all, or only within a set task?
Output	Does the child produce something, or only receive?

The fourth row separates most child-friendly tools from the rest. A tool that trains a model on the child's own examples has a bounded output. An open chat window does not, and that is precisely why the main chatbots set 13 as their limit.

What is allowed by age

The pattern is bounded tasks first, open chat much later.

Age	Suitable	Skip for now
6 to 8	Teachable Machine with a parent, drawing apps with AI features	Any chatbot, any account
9 to 10	Scratch with AI extensions, Code.org AI lessons	Personal chatbot accounts
11 to 12	Same, plus supervised chatbot use on a parent device	Using a chatbot alone
13 to 15	Own chatbot account with permission, history visible	Image generators without supervision
16 to 17	Most tools, with rules agreed in advance	Anything asking for payment details

An Indonesian family with a tablet at the dining table

Four tools worth starting with

All of them are bounded in scope, ask for almost no personal data, and make the child produce something.

Teachable Machine by Google. Your child trains a model with webcam photos in about ten minutes, then watches it fail when the examples are careless. No account, nothing stored permanently, and it explains how AI learns better than any explanation does
Scratch from MIT. Block coding with simple AI extensions. The account is optional and projects stay local until shared
Code.org. Free structured lessons including an AI section for primary ages, used in schools worldwide
Drawing apps with AI assistance, used beside you. Good for design-minded children, provided AI images are treated as drafts rather than finished work

Start with Teachable Machine this weekend. Ten minutes training a model to tell a pencil from an eraser gives more understanding than an hour of chatting with a bot, and it runs on any laptop with a camera.

Red flags worth refusing

Four signals mean the tool was not designed with children in mind.

No stated age limit, or an age buried in a long document
It asks for a photo of the child's face before anything works
It generates unlimited images from any text, with no visible filter
It pushes a subscription before showing what the free version does

The third is the one parents underestimate most. Open image generators are the fastest route to content nobody wanted, and they are rarely needed for anything a child actually requires. For broader safety guidance, Common Sense Media rates apps by age.

Settings to change on day one

Five minutes of setup removes most of the risk, and almost nobody does it.

Turn off training from chat history if the tool offers the option
Use a family email, not your child's school address
Disable microphone and camera access unless the tool genuinely needs them
Turn off image generation if the setting exists
Keep the device in a shared room, because that makes every other rule enforceable

Number five matters more than the other four combined. Rules that depend on trust collapse at ten at night; rules that depend on where the laptop physically sits do not.

What to teach alongside the tool

Safety with AI is mostly two habits rather than a list of banned apps.

Never type a full name, school, student number, address or phone number
Treat every answer as a draft that needs checking, especially numbers and sources

The first habit is easier to teach through one rule: retype the question, never paste the page. Children paste worksheet headers with their name and class without thinking, and that single habit removes the entire category of risk. The second matters because these tools sound equally confident when inventing, as we covered in our ChatGPT guide.

Safe is not the same as useful

A tool can be entirely safe and teach your child nothing. The difference is whether the child produces or only receives.

Type of tool	Safe	Teaches something
Model training, such as Teachable Machine	Yes	Yes, how AI learns from examples
Block coding with AI extensions	Yes	Yes, logic and sequencing
Homework answer apps	Mostly	No, the thinking is skipped
Open image generators	Depends	Rarely, without a task around it
Chatbots used to explain	With rules	Yes, when the child has to answer back

Row three is where most families end up by accident, because answer apps are the easiest to find and the quickest to install. They are not dangerous, simply hollow, and the marks gap shows up a term later.

An Indonesian parent and child at a laptop in the living room

A weekend setup plan

Saturday morning: run the five point check on whatever app your child already uses
Saturday midday: change the five settings above on every device
Saturday afternoon: try Teachable Machine together for ten minutes
Sunday: agree two rules out loud, no personal details and check every number
Sunday evening: ask your child to explain how the model they trained works

If your child can explain why their model got confused, the weekend worked. That single explanation covers more ground than any list of approved apps, because it means they understand AI learns from examples rather than simply knowing things.

When a class makes more sense

Supervision keeps a child safe; it does not make them capable. If your child has used these tools for months and still only receives answers without building anything, the gap is not the tool. Our AI classes for kids pair prompting with coding for ages 7 to 17, with the first class free.

Worth reading next: prompt writing for kids, is AI safe for children and what parents say.

Frequently asked questions

Which AI tools are safe for a primary school child?

The ones that give a bounded task rather than an open chat window. Teachable Machine by Google is the strongest starting point: the child trains a model using webcam photos, no account is needed, and in about ten minutes they see how AI learns from examples. Scratch from MIT adds simple AI extensions to block coding, with an optional account. Code.org provides free structured lessons including an AI section for primary ages. All three make the child produce something, which is the difference between a safe tool and a useful one.

How do I judge an AI app I have never heard of?

Five checks take about two minutes. Is there a stated age limit you can find in two clicks. Can it be used without a personal login for your child. Does it need photos, voice or personal details to function. Can it generate anything, or only work within a set task. Does the child produce something or only receive an answer. The fourth check separates most child-appropriate tools from the rest, because unbounded output is exactly why the major chatbots set an age of 13.

Can my child use ChatGPT?

OpenAI sets a limit of 13 and requires parental permission under 18, so below 13 there should be no personal account. At 11 to 12, supervised use on a parent's device is manageable if you are sitting there. From 13, an own account with permission is acceptable provided the chat history stays visible for the first few months. Under 13 and curious, the better route is a tool built for children, which teaches the same underlying ideas without an open conversation.

What are the red flags in an AI app?

Four are worth refusing outright. No stated age limit, or an age buried in a long document. A request for a photo of your child's face before anything works. Unlimited image generation from any text with no visible filter. A subscription pushed before the free version shows what it can do. The third is underestimated most often, because open image generators are the fastest route to content nobody wanted, and they are rarely needed for anything a child actually requires.

Which settings should I change first?

Five, and together they take about five minutes. Turn off training from chat history if the option exists. Use a family email, not your child's school address. Disable microphone and camera access unless the tool genuinely needs them. Turn off image generation if the setting is available. Keep the device in a shared room. The last has more effect than the other four combined, because rules that depend on trust collapse at ten at night, while rules that depend on where the laptop sits do not.

Are homework answer apps dangerous?

Not dangerous, more accurately hollow, which is a different problem. These apps are usually safe in data terms and teach nothing, because the child receives a finished answer and skips the thinking the exercise was designed to produce. Families end up there by accident, since such apps are the easiest to find and quickest to install. The cost appears a term later as a gap between homework marks and test marks. Exams in Indonesia are handwritten with no tool on the desk.

Can a child use an AI image generator?

With supervision from about 13, and with a task around it rather than complete freedom. The risk is not only unsuitable output; it also teaches a child that pictures simply appear rather than get made. Used inside a brief, such as producing three draft backgrounds for a project then editing one properly, it becomes a sensible design tool. Treat the output as a draft, check whether the platform sets an age limit, and do not let AI images be submitted as finished work without saying where they came from.

What should never be typed into any AI tool?

Full name, school name, student number, address, phone number, photographs and anything identifying other children. The practical version for a child is one habit rather than a long list: retype the question, never paste the page. Children paste entire worksheets, and the header carries a name and class without anyone thinking about it. The same applies to homework photos, which often include a school stamp or a teacher's handwriting. One retyping habit removes that whole category of risk.

How much screen time is reasonable for these tools?

Short and purposeful beats long and open-ended, and that matters more than any particular number. Ten to fifteen minutes with one goal, such as training a model or finishing one prompt exercise, leaves the child with something to show. An hour of open chatting usually leaves nothing. The practical rule is that the session ends when the task is done rather than when the time runs out, because that ties the tool to a purpose.

When is supervision alone not enough?

When your child has used these tools for months and still only receives answers without building anything. Safety and capability are two different problems, and settings only solve the first. If the pattern is passive, a class changes it by putting a project in front of the child and a teacher who watches the process. Our AI classes pair prompting with coding for ages 7 to 17, and the first class is free, which is enough to see whether your child starts making things or keeps waiting for output.

Blog

More from Blog

Algonova worldwide

Algonova Mexico Algonova Malaysia Algonova Global

Contact

Whatsapp: +62 21 50996971

Email: info@algonova.id

Address

Coworking Area, Gedung IT 6 Lantai Dasar, Nongsa Digital Park, Jalan Hang Lekiu, Kelurahan Sambau, Kecamatan Nongsa, Batam, Indonesia Directorate General of Consumer Protection and Trade Compliance – Ministry of Trade of the Republic of Indonesia. WhatsApp Ditjen PKTN: 0853-1111-1010.

Safe AI tools for kids: how to choose

The five point check

What is allowed by age

Four tools worth starting with

Red flags worth refusing

Settings to change on day one

What to teach alongside the tool

Safe is not the same as useful

A weekend setup plan

When a class makes more sense

Frequently asked questions

More from Blog

Online vs In-Person Coding Classes for Kids: Which Fits Better? (2026)

When Should Primary School Kids Start Coding: A Readiness Guide

8 Best Kids Coding Schools in Indonesia (2026): Review & Comparison

Four tools worth starting with

Red flags worth refusing

Settings to change on day one

What to teach alongside the tool

Safe is not the same as useful

A weekend setup plan

When a class makes more sense

Frequently asked questions

More from Blog

Online vs In-Person Coding Classes for Kids: Which Fits Better? (2026)

When Should Primary School Kids Start Coding: A Readiness Guide

8 Best Kids Coding Schools in Indonesia (2026): Review & Comparison