A sample chain of custody. Go ahead — tamper with it.
This is the record structure a regulator or opposing counsel would receive: the recommendation follows from the check, the check from the snapshot, the snapshot from the policy the AI actually read — each record sealed over the one before it. Every field below marked with a pencil is editable. Change anything. The chain re-verifies as you type, with real SHA-256, right here in your browser.
Source document — Investment Policy Statement, v3.2
Holdings snapshot — 41 positions
Concentration check — §4.2 limit test
Recommendation issued — human review sealed onto it
What you just did is what an insider, an attacker — or an AI rewriting its own story — would try. In the product, the chain above is only the first plane: records are mirrored to an independent append-only store, cross-checked continuously, and anchored to external timestamp authorities (RFC 3161), so even the operator of the database cannot silently rewrite history. Your auditor verifies independently, without contacting us. In the live demonstration, we do this against the real product.