How do I prove a dataset has not been tampered with?

Seal it: hash every record, then hash those hashes into a single root. Anyone re-running the same SHA-256 on the data reproduces the root exactly — if any value changed, the root changes, and comparing the per-record hashes identifies which record moved. The demo on this page does exactly that, live in your browser, with nothing uploaded.

Can you verify data without seeing the data?

Yes. The certificate is a list of hashes and contains zero business data. A hash reveals nothing about the values behind it, so the certificate can be handed to an auditor, a regulator or a customer without exposing a single record — while still proving the records are exactly what was sealed.

Does tamper detection tell you which record changed?

It should, and AT-1's does. A plain file checksum only tells you that something changed somewhere. Because each record is hashed individually, verification pinpoints the specific record that no longer matches and proves the remaining records are untouched — the difference between an alarm and usable evidence.

Do I need AT-1 software to verify an AT-1 seal?

No. Verification is plain SHA-256, a public standard implemented in every language and in every browser. This page proves the point by verifying with no backend, no account and no AT-1 code — you can read the source. Proof that depends on trusting the vendor is not proof.

Is tamper-evidence the hard part of verifiable data?

No, and we say so. Tamper-evidence is well-understood cryptography, which is precisely why we let you check it yourself in ten seconds. The hard problems sit underneath: querying data while it stays sealed and compressed, and proving a record was genuinely deleted without disturbing the records around it.

Verify it yourself

Change one byte.

This fuel-retail table is sealed. Click any value and change it — the seal will catch you, and point at exactly which record moved.

Real SHA-256, computed in your browser. Nothing is uploaded. No account. View source if you don’t believe it.

SEAL INTACTAll 12 records match the certificate.

Sealed dataset — site 04, dispenser readingsclick a value to alter it

Record	Timestamp	Pump	Product	Litres	Tank level

Certificate

Root hash (recomputed live): —
Sealed root (as issued): —
Verdict: VERIFIED — root hash reproduces exactly.
Payload: 0 bytes of business data. Hashes only.

It pinpoints, it doesn't just flag

Most integrity checks tell you a file changed. This tells you which record — so you can act on it instead of re-auditing everything.

The certificate holds no data

What proves the table is a short list of hashes. Hand it to an auditor, a regulator or a customer without exposing a single reading.

Anyone can check it

Verification is plain SHA-256 — no AT-1 software, no licence, no trust in us. This page proves it by working without a server.

This is the easy layer, and we’ll say so

Tamper-evidence is well-understood cryptography — we lead with it because it is the part you can verify yourself in ten seconds. The hard problems are what sit underneath: querying data while it stays sealed and compressed, and proving a record was truly deleted without disturbing the rest.

See it on a customer’s production data Provable deletion Query it while it stays sealed The same idea, applied to model weights

Common questions

How do I prove a dataset has not been tampered with?: Seal it: hash every record, then hash those hashes into a single root. Anyone re-running the same SHA-256 on the data reproduces the root exactly — if any value changed, the root changes, and comparing the per-record hashes identifies which record moved. The demo on this page does exactly that, live in your browser, with nothing uploaded.
Can you verify data without seeing the data?: Yes. The certificate is a list of hashes and contains zero business data. A hash reveals nothing about the values behind it, so the certificate can be handed to an auditor, a regulator or a customer without exposing a single record — while still proving the records are exactly what was sealed.
Does tamper detection tell you which record changed?: It should, and AT-1's does. A plain file checksum only tells you that something changed somewhere. Because each record is hashed individually, verification pinpoints the specific record that no longer matches and proves the remaining records are untouched — the difference between an alarm and usable evidence.
Do I need AT-1 software to verify an AT-1 seal?: No. Verification is plain SHA-256, a public standard implemented in every language and in every browser. This page proves the point by verifying with no backend, no account and no AT-1 code — you can read the source. Proof that depends on trusting the vendor is not proof.
Is tamper-evidence the hard part of verifiable data?: No, and we say so. Tamper-evidence is well-understood cryptography, which is precisely why we let you check it yourself in ten seconds. The hard problems sit underneath: querying data while it stays sealed and compressed, and proving a record was genuinely deleted without disturbing the records around it.

Sample data is synthetic and stands in for a real wetstock feed. The seal, the hashes and the tamper detection on this page are genuine and run entirely client-side.