Guide
How BrainWave uses AI safely: what it sees, what it does not, and who decides
BrainWave uses AI to help you, the parent, draft lessons and quizzes. Drafting never uses your child’s name, answers, scores, points or progress. Everything it writes waits for you to read and accept it, so nothing reaches your child that you have not approved. The one exception is Ask Ria & Ravi, a feature your child can use to ask a question about a lesson, which is off until you switch it on and is covered in its own section below.
That is the short version. The rest of this guide explains what the AI does receive, what it cannot do, where the limits are, and what you should still check yourself. AI is useful for saving time, but it makes mistakes, and a parent who reads the draft is the real safeguard.
What the AI receives
When you ask for a lesson or quiz, the AI gets three things.
- The lesson text you pasted, imported or are working on.
- The instruction you typed in the chat.
- Any photo or PDF you attached, for example a worksheet.
Photos are shrunk on your phone and re-encoded before they are sent, which drops the photo’s location (GPS) data before it leaves the phone.
What drafting never receives
No part of a child’s record is put into a drafting request. That means no name, no answers, no scores, no points and no progress. Your child does not even have an account: there is no email, password, birth date or photo for the app to hold.
This is a design choice. The AI writes about the lesson, not about your child. It cannot tailor a question to your child’s earlier mistakes because it does not know them, and we think that is a fair trade.
There is one place where your child’s record does reach the AI, and only when you ask: the parent assistant. If you ask it something like “how did my son do on fractions this week?”, it looks up your own household’s records with fixed, read-only lookups and passes the results, which can include your children’s display names, scores and points, to the AI so it can write the answer. It cannot see other households and cannot change anything. If you would rather it never sees those details, check progress on the child’s page instead of asking the assistant.
Ask Ria & Ravi: when your child asks the AI
Ask Ria & Ravi lets a child type a question about a lesson and get a short answer from Ria or Ravi. It is off by default; you switch it on for each child.
- What the AI receives: only the question, the lesson text and your child’s reading age. No name, no scores, no earlier questions. It always runs on BrainWave AI (Claude by Anthropic), never your own key, and Anthropic does not train on it.
- Short answers, in words suited to your child’s reading age.
- Private details: a question containing an email address or phone number is not sent to the AI.
- Worrying questions: if a question suggests your child is hurt or unsafe, the AI does not answer. Your child is told to talk to a grown-up they trust (in India, Childline 1098 is free any time) and you are notified.
- Grown-up topics are gently redirected to you.
- You see everything: every question and answer. You can report an answer and delete questions. The limit is 30 questions per child per day.
Unlike a draft, an answer reaches your child straight away and you see it afterwards, which is why it is off until you choose to turn it on. Questions are deleted when you delete them or delete your child.
You are the gatekeeper
AI output is a draft. In BrainWave:
- Quizzes are drafted by AI, and you keep, edit, delete, write your own or ask for more.
- Ria and Ravi stories are short animated stories made from a lesson. You read and approve the words before your child can see them.
- Nothing is published to the child until you publish it.
Do read the drafts closely. The AI can state something wrong with total confidence, or write a question with two correct answers. Our guide on turning a worksheet photo into a quiz has a checklist for reviewing a draft.
Reporting a problem
If the AI writes something wrong or unsuitable, use the Report this AI content option. You can also simply delete the draft and start again with a clearer instruction. Reporting helps us see problems we would not otherwise hear about.
Which AI, and whose key
BrainWave AI is included free and runs on Claude, made by Anthropic. Each household has a monthly allowance, and Settings shows the percentage used and the renewal date, which is the 1st of the month.
If the allowance runs out, BrainWave uses your own key if you added one. Otherwise it tells you the allowance renews on the 1st. You may optionally add your own API keys from Anthropic, OpenRouter, Groq or Cerebras, choose a model for each key, and decide whether BrainWave AI or your own keys go first. Google Gemini is not offered.
Using your own key is optional. Many families never need to.
What happens to your content
Sazinga does not use content to train any model. Lessons, quizzes, names and progress are stored on Sazinga’s server in plain text, not encrypted so that only you can read it, which means Sazinga can technically read them. They are not sold or shared.
We say that plainly because it matters: if you are not comfortable with a piece of text being on a server, do not paste it. For example, do not put a school report with your child’s name into a lesson. You can delete everything from the app under More, then Delete account.
There are no adverts, no subscription and no in-app purchases, and no analytics or tracking on any screen a child sees.
What the AI does not do
- It does not chat with your child, apart from Ask Ria & Ravi when you have switched it on.
- It does not grade your child’s work. Quizzes mark themselves from the answers you approved.
- It does not decide rewards. You set and price every reward, and every redemption waits for your approval.
- It does not replace a teacher. It cannot assess your child’s reading or spot a learning difficulty. If you are worried, talk to their teacher or a specialist.
Sensible habits for parents
- Read every question and every story line before you publish.
- Keep names and private details out of what you paste or photograph.
- Check any fact you are not sure about.
- Report anything that looks wrong or unsuitable.
- Keep screen use short and balanced with real books.
The full details are on the AI page, and the privacy terms are on the privacy page.
Frequently asked questions
Does the AI ever see my child's name or scores?
Not when it drafts. A lesson or quiz request carries the lesson text, your instruction and any photo or PDF you attach, and nothing from your child's record. The parent assistant is the exception: if you ask it about your family, it looks up your own household's records, which can include names and scores, to answer you.
Can the AI show something to my child without my approval?
No. Nothing the AI writes reaches a child until you have reviewed and accepted it. Lessons, quizzes and Ria and Ravi stories all wait for you.
Which AI does BrainWave use?
BrainWave AI runs on Claude, made by Anthropic, and is included free. You can optionally add your own API keys from Anthropic, OpenRouter, Groq or Cerebras. Google Gemini is not offered.
Is my content used to train AI models?
Sazinga does not use your content to train any model. Lessons and quizzes are stored on Sazinga's server in plain text, so Sazinga can technically read them, but they are not sold or shared.
Can my child ask the AI questions?
Only if you switch on Ask Ria & Ravi for that child. It is off by default. The AI receives just the question, the lesson text and the reading age, you can see every question and answer, and a worrying question is not answered by the AI but passed to you.
What if the AI writes something wrong or unsuitable?
Use the Report this AI content option to tell us. Because you review everything before your child sees it, you can also delete or rewrite it yourself.