Define the question
Pick one small classification task, such as ripe/not ripe, rolls/does not roll, or animal/vehicle.
Parent decision table
Choose the lowest-data tool that teaches the idea
| Tool or method | Best fit | Age guidance | Data and privacy boundary | First safe activity |
|---|---|---|---|---|
| Offline sorting classifier | Understanding training examples, labels, testing, and failure. | 7–14 with an adult. | No account, camera, microphone, upload, or child profile. | Sort household objects into two labels, test hidden examples, and record mistakes. |
| Scratch | Visual coding, interactive stories, games, and simple AI-related logic. | Scratch says it is primarily designed for ages 8–16 and can be used by younger children with parents. | Use offline or parent-reviewed sharing; avoid names, school, contact details, and identifying media. | Create a sprite that asks questions and explains why it chose one branch. |
| Google Teachable Machine | Seeing image, sound, or pose classification from examples. | Parent-guided; review Google policies before webcam, microphone, or account use. | Google says training can happen locally on the device; still avoid child faces, voices, bedrooms, school logos, and private files. | Train “object A vs. object B” with toys or drawings, then test confusing examples. |
| CodeAI / Code.org AI tools | Classroom AI lessons, teacher-led AI chat, and curriculum-aligned activities. | Some AI Chat Lab access is limited to verified-teacher sections and is not available to students under 11. | Use only through authorized teacher, school, district, or parent setup; avoid free-text personal details. | Use a public AI lesson first, then discuss what the tool should and should not answer. |
| micro:bit and MakeCode | Connecting code, sensors, movement, and simple physical computing. | micro:bit resources are designed for children and educators around ages 8–14. | Movement samples in micro:bit CreateAI can remain in browser local storage; downloaded files may contain sample data. | Use the simulator or a physical micro:bit to classify simple gestures without names or photos. |
SAFE-BUILD-EXPLAIN routine1 2 3 4
A repeatable training-tool lesson
Choose safe examples
Use objects, drawings, public sample data, or invented text. Do not use child faces, voices, names, schoolwork, or locations.
Train and test
Keep some examples hidden until testing. Record wrong answers instead of hiding them.
Explain the mistake
Ask which feature, label, or missing example caused the model to fail.
Before a child uses an AI training tool
Parent safety checklist
- Confirm the tool’s current age rules, classroom rules, account requirements, and data-use policy.
- Use adult-operated accounts when a child is under the provider’s independent-use age.
- Prefer no-account, offline, or local-browser activities before camera, microphone, or uploaded files.
- Use non-identifying examples: toys, drawings, objects, public-domain images, or synthetic text.
- Require a short reflection: what did the model learn, where did it fail, and who should check it?
Sources checked September 2026Request a Free Class
Primary sources to review before sign-up
- Scratch for Parents — age range, community, offline use, and privacy overview.
- Google Teachable Machine and Teachable Machine FAQ — model types and local-device training explanation.
- CodeAI AI Chat Lab FAQ and CodeAI Privacy Policy — teacher-led access, under-13 requirements, and privacy policy.
- micro:bit privacy policy, micro:bit terms, and MakeCode for micro:bit overview — age audience, personal-information guidance, and editor details.
Want help choosing the first tool?
Use the free class to start with AI basics, privacy boundaries, and a parent-guided demonstration.