AI

How BrainWave uses AI

AI in BrainWave works for the parent: it drafts lessons and answers your questions. A child can ask it a question about a lesson only if you switch that on. It never publishes anything and never decides who earns what.

Included, on by default

BrainWave AI runs on Claude, made by Anthropic. Every household gets it free, with a monthly allowance and nothing to set up. Settings shows how much of this month's allowance is used, as a percentage, and when it renews on the 1st.

The allowance is there to stop abuse, not to ration a family's reading. If it does run out, BrainWave moves to your own key if you have added one; otherwise drafting pauses until the 1st, and everything else in the app keeps working.

What it does

  • Drafts a lesson from your description, a pasted text, up to six photos of a worksheet or textbook page, or a PDF.
  • Writes multiple-choice questions from a lesson, at the reading level you set, spread across the whole text.
  • Writes the script of a Ria and Ravi story from a lesson.

Each of these is a draft on your screen. You edit it, keep it or bin it. Nothing reaches your child until you publish it.

What reaches the AI, and what never does

When you…The AI receivesIt does not receive
Draft a lesson, quiz or story The lesson text, the instruction you typed, and any photo or PDF you attached Anything from your child's record: name, answers, scores, points, rewards or progress
Your child asks a question in Ask Ria & Ravi The question the child typed, the text of the lesson, and the child's reading age The child's name, scores, points or any earlier question. Emails and phone numbers in a question are not sent at all
Ask the parent assistant about your family, such as "how did she do this week?" Your question, and the results of read-only lookups in your own household, which can include your children's display names, scores, points, tasks and rewards Anything from another household. The lookups cannot change anything, and run only when you ask

Photos are shrunk and re-encoded on your phone before they are sent, which also strips the location data a camera stores in them. Your child's own activity is never sent to an AI by itself. An AI call happens when a parent asks for one, or when a child asks a question in Ask Ria & Ravi, which a parent has to switch on first.

Ask Ria & Ravi: a child's question

While reading a lesson, a child can type a question and get a short answer from Ria or Ravi. It is off until you switch it on, separately for each child, in the parent app.

  • BrainWave AI only. Claude by Anthropic answers. Your own keys are never used for a child's question.
  • Short and age-shaped. Answers are a few sentences, written for the child's reading age and tied to the lesson.
  • Private details stay out. A question with an email address or a phone number is not sent to the AI; the child is asked to keep that private.
  • Worrying questions go to a person. If a child asks something that suggests they are hurt or unsafe, the AI does not answer. The child is told to talk to a grown-up they trust (in India, Childline 1098 is free at any time) and you are notified.
  • Grown-up topics are redirected. The child is gently pointed to you, and you are told.
  • You see everything. Every question and answer is visible to you. You can report an answer and delete questions.
  • A daily limit. 30 questions per child per day.

Only the question, the lesson text and the reading age are sent. Anthropic does not train its models on it. Questions are deleted when you delete them or close your account.

Your own keys, if you want them

You can add API keys from Anthropic, OpenRouter, Groq or Cerebras, pick a model for each, and choose whether BrainWave AI or your own keys go first. Keys are stored encrypted. Requests made with your key are billed by that provider, under its terms.

Google Gemini is no longer offered. Its API terms do not allow apps likely to be used by under-18s, so it was removed and existing Gemini keys were switched off.

What it does not do

It does not chat with your child except through Ask Ria & Ravi, and only when you have switched that on. It does not mark quizzes; the quiz marks itself against the answers you approved. It does not award or spend points. It is not used to train any model by Sazinga, and it can be wrong, which is why you see every draft first. The guide to using AI safely covers how to check a draft in a couple of minutes.

Questions about AI

Which AI does BrainWave use?

BrainWave AI runs on Claude, made by Anthropic. It is included free for every household, with a monthly allowance. You do not need an account with Anthropic or a key of your own.

How big is the monthly allowance?

Large enough for everyday use: drafting lessons and quizzes for a family's reading. Settings shows how much of this month's allowance is used, as a percentage, and the date it renews, which is the 1st of the month.

What happens when the allowance is used up?

If you have added your own key, BrainWave switches to it. If you have not, AI drafting pauses with a message saying when the allowance renews. Lessons, quizzes, stories and rewards keep working, and you can still write anything yourself.

Which of my own keys can I add?

Keys from Anthropic, OpenRouter, Groq or Cerebras. You choose a model for each key, and whether BrainWave AI or your own keys are tried first. What your own key costs is between you and that provider.

Why is Google Gemini not offered?

The Gemini API terms do not allow it in apps that are likely to be used by people under 18. BrainWave's child app is exactly that, so Gemini was removed and existing Gemini keys were switched off.

Does the AI ever see my child's name or scores?

Not when it drafts. A lesson, quiz or story request carries the lesson text, your instruction and any photo or PDF, and nothing from a child's record. The parent assistant is different: when you ask it about your family, it looks up your own household's records, which can include names, scores and points, so it can answer you.

Is my content used to train AI models?

Sazinga does not use your lessons or your children's activity to train any model, and does not sell or share them. If you use your own key, that provider's terms apply to the requests made with it.

What if the AI writes something wrong or unsuitable?

You see every draft before your child does, and you can edit or delete it. If a draft is wrong or inappropriate, use Report this AI content on the chat or draft screen, which sends the job and your reason to us.