How to choose safe technology for your child: a 10-point checklist

Before the checklist: the one-sentence test

Every children's product has a business model, and the business model predicts behaviour better than the marketing does. Ask: how does this company make money? Good answers are boring — you pay for the product, or a school does. Bad answers hide: free products funded by advertising, data, or in-app purchases have a structural need for your child's attention and information. If you cannot find the answer in two minutes, treat that as the answer.

The 10-point checklist

One: data collection — does it collect only what the feature needs? A drawing app requesting your contacts fails. Two: retention and deletion — can you see and delete your child's data yourself, without emailing support? Three: training and sharing — is data sold, shared, or used to train AI models, and is that off by default? Four: ads — are there any, and are they mixed into content? Advertising to young children fails outright. Five: purchases — can a child spend money or unlock purchases without a parent step-up?

Six: contact — can strangers reach your child (chat, comments, friend requests)? Any open contact channel needs a hard look. Seven: dark patterns — streaks, countdown timers, 'your friend is waiting', endless autoplay: these are engineered compulsion, and in a children's product they tell you whose side the designer is on. Eight: content controls — can you scope what the product will discuss or show, by age and topic? Nine: parent visibility — do you see what the company sees, or only a comforting subset? Ten: exit cost — can you export or wipe everything and leave cleanly?

Score honestly: eight or more is a product you can live with; five to seven means enable every control and re-check monthly; below five, walk away — there is always an alternative.

Reading a privacy policy in three minutes

Search the policy for five words: 'share', 'third part', 'retain', 'advertis', and 'train'. Read only the sentences containing them. This ten-minute-looking task takes three and surfaces ninety percent of what matters. A policy that is short, specific, and written in human language is itself a signal; one that reserves the right to change terms silently is too.

For AI products specifically, add the questions from our companion guide: where does audio go, what is stored, and what stands between the model and your child.

Our own scorecard

We publish RAFEEQI's answers to all ten points on our safety page, and we will keep them current as the product evolves from scripted demo to live companion. The checklist above is the test we expect parents to run against us — please do, and tell us where we fall short. A safety promise that fears the checklist is not a safety promise.