Every record carries a signed consent receipt

Your data is already training AI. Get paid for it.

Datavar turns the data you already produce into a monthly payout, and gives AI teams the one thing scraped data can never have: consent.

Become a contributorLicense datasets

Built by people previously at

For people

You produce the data.
You should set the terms.

Right now your data is taken quietly and monetised by someone else. Datavar flips the default: nothing moves until you say so, and when it moves, you're paid.

  1. 01

    Connect what you already have

    Link the accounts, devices and files that hold your data. Access is read-only, scoped, and revocable the second you change your mind.

    • Browsing & search
    • Health & wearables
    • Purchases
    • Media
  2. 02

    Nothing leaves until you approve it

    You approve each dataset, each buyer, and each purpose separately. Every approval is written as a signed consent receipt with an expiry date.

    • Per-buyer control
    • Purpose limits
    • Expiring by default
  3. 03

    Get paid when it sells

    Rare, high-signal data earns more than common data. Payouts settle in XLM straight to your wallet, with the transaction hash to check them against.

    • XLM payouts
    • On-chain receipts
    • PayPal soon

Earnings

See what you're currently giving away for free.

Pick the sources you'd be willing to share. Nothing has been licensed on testnet yet, so there are no averages to quote. These fill in from real sales as the protocol runs.

Your sources

1×

Paid out as

Estimated payout

—

no settled sales to average yet

  • Browsing & search—
  • Purchase history—
  • Streaming & media—

Rare data earns more than common data. A dashcam clip from a rainy roundabout is worth more to a robotics team than another hour of scrolling.

Claim your payout

For AI teams

Scraped data is a lawsuit with a training loss.

Every record in Datavar is traceable to a consenting human, a signed purpose, and an expiry date. Filter a cohort, inspect the distribution, license it through one API.

  • Provenance on every row

    Each record resolves to a consenting person, an agreed purpose, and an expiry. Your legal team can audit any row you trained on.

  • Recruit the cohort you actually need

    Filter by source, region, device, or demographic, then commission data that doesn't exist yet. Contributors opt in and start producing in days.

  • Fresh, not frozen

    Scraped corpora are a snapshot of a dead internet. Datavar streams continuously, so your model sees this month, not 2023.

  • Licensed and indemnified

    Commercial terms, warranties, and indemnity on every dataset. Deletion requests propagate to you within 24 hours.

cohort.py
# 4,102 consenting drivers, EU, streaming live
from datavar import Client

client = Client(api_key="gk_live_…")

cohort = client.cohorts.create(
    sources=["dashcam", "location"],
    region="EU",
    consent="commercial",
    fresher_than="30d",
)

for record in cohort.stream():
    train(record.payload)
    audit(record.receipt) # signed, revocable

Docs coming with the buyer release

The protocol, so far

Testnet · live

1 square = 1 dataset · accent = licensed at least once

contributors
2
datasets contributed
3
paid out to date
0 XLM
payouts settled on-chain
0

FAQ

The questions you should be asking.

If something here reads like a dodge, tell us and we'll rewrite it.

Are you selling my identity?

No. Identity is stripped when data is ingested and buyers receive pseudonymous records. They can license the fact that a 30-year-old in Berlin walked 8,000 steps; they cannot license you. Re-identification is contractually forbidden and technically monitored.

How much will I actually earn?

Most contributors land between $20 and $60 a month with three or four sources connected. Rare data pays far more: dashcam footage, clinical wearables, and low-resource languages are the current premiums. Anyone promising you thousands is lying.

What happens if I change my mind?

Revoke any consent from your dashboard and it stops immediately. Buyers are notified within 24 hours and are contractually required to delete the affected records. Data already used to train a shipped model can't be un-trained, which is why consent is scoped and expiring by default.

Do I need a crypto wallet?

For now, yes. Your wallet is your account — there's no separate login — and payouts settle in XLM on Stellar, which is also what makes them cheap enough to pay out per sale. PayPal is planned so you won't have to, but it isn't live yet.

How is this different from a data broker?

A broker takes data you never knowingly gave and sells it without paying you. We invert all three: you choose what leaves, you see who buys it, and you take a cut of every sale. If that sounds like a worse business, it is. For the broker.

For buyers: why not just scrape?

Because you can't audit it, can't refresh it, can't commission what doesn't exist yet, and increasingly can't defend it. Datavar costs more per record than scraping and considerably less than discovery.