NoteSparkAIYour notes, uploads, and AI conversations are never used to train any AI model — ours or our providers’. We enforce that with contractual zero-retention agreements, API settings that opt out of training, account-scoped data access, and twice-yearly audits.
I’m the CEO, so I’ll be direct about why you should care and exactly what we do. A promise without a mechanism is marketing. Here’s the mechanism.
Why this matters more for study notes than almost anything
Your notes are an unusually intimate dataset. They reveal what you don’t understand yet, what you’re struggling with, the half-formed ideas you’d never publish. For a model trained on the open web, that’s exactly the kind of private, high-signal text that should never leak into a training set where it might one day surface in someone else’s output.
“We don’t sell your data” is the floor, and frankly it’s a low one. The question that actually matters for an AI product is narrower: is my content used to train models?For us the answer is no — and the rest of this post is how we make that true rather than merely stated.
The four things that make the promise real
- Zero-retention agreements with model providers.When your text is sent to a model for inference, it’s processed and discarded — not stored, not retained, not used for training. This is a contractual term with every AI provider we use, not a default we hope holds.
- Training opt-out at the API level.Enterprise AI APIs let customers disable any use of their data for model improvement. We use those settings on every request. Your content goes in, an answer comes out, nothing about it is kept to make the model “better.”
- Account-scoped access. Your library is retrievable only for your answers. The citation engine never reaches into another user’s notes, and we don’t pool content into a shared index.
- We don’t train our own models on you either.The easy loophole would be “providers don’t train on it, but we do.” We don’t. Your content isn’t a training corpus for any first-party model.
We’ll show the relevant contractual terms to institutions and security teams that ask. A promise you can’t inspect isn’t worth much.
Trust, but verify — including ourselves
Internal rules drift unless someone checks. Twice a year we audit our data flows: which services receive your content, what their retention terms say, and whether our code actually matches the promise on this page. When a provider changes its terms, that review is how we catch it — and if a provider ever stopped offering zero retention, we’d move off it rather than quietly weaken the guarantee.
You hold the other half of the controls. You can export everything you’ve created as open JSON, and you can delete your account and its contentswhenever you want — content is purged from primary storage within 24 hours and from backups within 30 days. No retention dark patterns, no “are you sure” maze.
The bargain we actually want
Plenty of free products are free because you’rethe product — your data is the thing being sold or mined. We deliberately chose a different deal: paying subscribers fund the model bills, so free users don’t have to pay with their privacy. That’s only possible if “we don’t monetize your data” is a real constraint, not a slogan we’d abandon under pressure.
Studying means writing down what you don’t know yet. You should be able to do that without wondering where it ends up. That’s the whole promise — and now you know how it’s enforced.
Read the full Privacy Policy, or start free— your notes stay yours.
Frequently asked questions
Does NoteSparkAI use my notes to train AI models?
No. Your notes, uploads, and AI conversations are never used to train any model — ours or our providers'. Inference happens under contractual zero-retention agreements, and we opt out of data-for-training settings on every API request.
Can I delete my data?
Yes. You can export everything as open JSON and delete your account at any time. Content is purged from primary storage within 24 hours and from backups within 30 days.
How do I know the no-training promise is real?
It's backed by contractual terms with model providers, API-level training opt-outs, account-scoped data access, and twice-yearly internal audits of our data flows. We'll share the relevant terms with institutions and security teams on request.