For AI assurance providers: evidence you can test yourself.
You are asked to form a view on what an AI system did. The records you are shown are produced by the organisation you are assessing. Sigilbase gives you a way to test them that does not depend on their word, or ours.
What you receive
An evidence bundle: the events for a period, the signed checkpoints that seal them, the public keys and a copy of the verifier. One archive, no account, nothing to install beyond the verifier's runtime.
What the verifier checks
Every event hashes to the value the chain records and links to the one before; every checkpoint's Merkle root is recomputed from its events; every checkpoint is signed by the published key; checkpoints form an unbroken chain; where anchoring is on, each checkpoint root was timestamped by an independent authority. A changed, deleted or reordered event fails at that record, by sequence number.
What it tells you, and what it does not
It tells you the records have not changed since Sigilbase received them, and when that was. It does not tell you the records are complete, that the model behaved as the records say, or anything about the period before the organisation started sending events. Those are questions for your other procedures.
Working with a client who uses Sigilbase
Ask for auditor access. It gives you read access to named streams and a date range for your engagement, up to 90 days at a time, through one-time sign-in links. What you open is recorded in the client's ledger, which protects both of you. Export what you need, where the client's grant allows it. Keep the bundle; it verifies on its own, for as long as you keep it, whether or not Sigilbase exists.
The auditor guide covers signing in and what to do when you meet a redacted event.
Get in touch
We write for assessors and we take correction. If a claim on this site overreaches, tell us and we will change it. hello@sigilbase.io.
What this proves, and what it does not
Sigilbase proves that a record has not been modified, deleted or reordered since we received it, when we received it, and which identity sent it. Any lawful redaction is declared, never silent.
It does not prove that a model produced the output in the record, or ran at all. It does not prove a decision was accurate, fair or lawful. It does not prove that everything which happened was recorded; coverage is the sender's control. It does not prove the sender's claimed time, only ours. And it says nothing about the period before the sender started.
Sigilbase is the evidence layer. The conclusion belongs to the person examining the evidence.
Start recording provable history
Chained, sealed, independently verifiable audit logs, from the first event. Free while Sigilbase is in beta.
Start free Download the verifier
Questions first? Write to hello@sigilbase.io.
More in this section
- AI audit records. Why an AI decision log needs a custodian, and the rules that ask for one.
- AI decision logging. One event per decision, what to put in it, and what it proves.
- Records of what an AI agent did. Agents as actors, tool calls as events, credentials on loan.
- EU AI Act record-keeping, plainly. Articles 12, 19 and 26 in plain words, and where Sigilbase fits.
- UK automated decisions. The 2025 Act's safeguards, and the record a contested decision needs.
- ISO/IEC 42001 event logs. What control A.6.2.8 asks for and what a certifier samples.