There's a new habit quietly spreading: people pasting their bank statements into a general AI chatbot and asking, "where is my money going?"
It makes sense. The answers are good. But here's the part most people don't stop to ask: what exactly did I just hand over, and where does it live now?
At Centiv, an AI helps with your money too — but it's built on a different promise. The short version: it sees your spending, not your identity. Here's exactly what that means, in plain terms.
What the AI never sees
Before a single question reaches our AI, every personal identifier is stripped out. Not hidden — removed. The following never appear in any prompt:
- Your name, email, or phone number
- Your full bank account number (only the last-4 mask shows, like on a paper statement)
- Your routing number, date of birth, or address
- Your Social Security number (we don't collect it in the first place)
- Your bank login, access tokens, or any authentication secret
- Your Centiv user ID, IP address, or device identifiers
The result is simple: even if someone could read every prompt our AI has ever received, there would be no way to know whose money they were looking at. It's a stream of financial behavior with no person attached.
What it does see — and why
To answer "can I afford this?" or "where did my paycheck go?", the AI does need your actual financial picture. So it receives a scoped snapshot:
- Merchant names your bank reports (e.g. "Starbucks", "Whole Foods")
- Transaction dates and amounts
- Account types and balances (e.g. "Checking · $2,431")
- Recurring subscriptions and their cost
- Your ¢Score and recent changes
- Your stated focus (like "build savings") if you've set one
None of that, alone or combined, identifies you. It's enough for an advisor to reason about your money — and useless for tying it back to a person.
One honest caveat: if you upload a receipt, the image is passed to the model as-is so it can read the line items. If your receipt has your name printed on it, that's in the image. We surface this so you can choose — you can always describe a receipt in text instead.
Your data never trains a public model
Centiv has a Data Processing Agreement with our AI provider. Under it, anything we send is used only for your request and is never used to train any public model. That's a binding contract, not a toggle that can quietly flip.
When you type your finances into a general consumer chatbot, you're governed by that product's consumer terms — which often include training defaults that shift over time, sometimes with opt-outs you have to go find yourself.
Read-only by design
We connect to your bank through a trusted bank-link partner — the same trust layer used by most major US fintech apps — and we request read access only.
That means Centiv physically cannot move your money. We can't initiate a transfer even by mistake, because the integration was never granted that capability. Even in a worst-case security failure, an attacker couldn't use our connection to drain your account. Your bank password is never seen by Centiv either — it goes straight from you to your bank.
Six focused advisors, not one open chatbot
Centiv isn't a single all-knowing bot. It's six purpose-built advisors — Coach, Sage, Ace, Warden, Sentry, and Clerk — each with a focused job and an audited prompt. Each gets only the data its job needs. None of them keep a long-term, cross-user memory of who you are.
You can verify it — and take it all back
Trust shouldn't require faith. So:
- Every internal data read is logged with a reason, and you can see it on your phone under Settings → Privacy → "Who's accessed my data." Most fintech companies don't show users this. We do.
- Export everything — accounts, transactions, ¢Score history, subscriptions — as JSON or CSV, anytime.
- Delete your account in a tap; records leave active systems within 7 days, backups within 30.
- Disconnect any bank instantly, and we stop receiving updates.
And the promises are enforced in code, not just policy: a test runs on every change and fails the build if any personal identifier slips into an AI prompt. We'd rather ship a feature a week late than ship a privacy regression.
The bottom line
An AI can be genuinely helpful with your money and respect your privacy — but only if it's built that way on purpose. Centiv strips your identity, signs a no-training contract, stays read-only, and shows you the receipts.
Want the full, provable version? Read how Centiv handles your data and AI, or the formal Privacy Policy.