SYNTHETIC DEMO — NO REAL CUSTOMER DATA
AIDALEvidence specification ↗

Can you prove what your AI
actually decided?

When an automated decision is challenged, “it’s in our database” isn’t evidence — you control the database. This is a real evidence record, verified in your browser.

AIDAL evidence case
Loan application #48291
aud_59465349a3512931dd1850e8
TAMPERED
Decision
DECLINED
Evidence
TAMPERED
Privacy
DIGEST ONLY
Compliance
NOT ASSESSED
The product
1
AI makes a decision
2
AIDAL seals what was submitted
3
The decision is challenged
4
Compliance investigates
5
Evidence is independently verifiable
Without AIDAL

“Show us exactly what happened.”

Engineering searches the database, application logs, model logs, cloud logs, backups, model versions, internal docs — then says: “we believe this is the record.”
With AIDAL

“Show us exactly what happened.”

Compliance finds the decision, verifies it, exports the package, hands it over. No engineering ticket.
Why not build this yourself?

You can build the hash chain. You probably should, if all you need is internal integrity.

The reason to use AIDAL is everything around it: canonicalization, schema versioning, offline verification tooling, evidence exports, credential attribution, regulatory mapping, adversarial testing, backward compatibility with years-old records, and documented limitations your auditor can read.

AIDAL isn’t selling SHA-256. We’re selling an evidence workflow your compliance team can actually use.

Don’t take our word for it
Evidence specification ↗Threat model ↗Canonical format ↗Test vectors ↗Offline verifier ↗Public transparency ↗

We publish the assumptions behind our own evidence system — including where it is weakest.

Get an API keyanthony@tryaidal.com

How much time does your team currently spend reconstructing a challenged AI decision? That’s the number worth measuring — we won’t invent one for you.