Evidence packs
LiveFreeze a risk report and hash it, so you can prove later what the product said at the time.
- ✓Frozen risk-report snapshot with a content hash
- ✓Verification that re-hashes and compares
- ✓No edit and no delete — by design, not by omission
- ✓Repeat freezes allowed, so a later pack can prove an earlier one is unchanged
What it is.
An evidence pack is a frozen copy of a contract's risk report at one moment, with a hash over its contents. Later, a verify call re-hashes the stored payload and tells you whether it still matches — so a pack is a record you can stand behind rather than a screenshot.
There is no way to edit or delete a pack, and that is the feature rather than an omission: a record you can change after the fact answers nothing. A corrected analysis is a new pack, and both remain. Freezing the same report twice is allowed on purpose — the second pack proves the first was not altered in between.
Three steps.
End to end.
Create a pack from a contract's current risk report.
Packs are listed newest first. Nothing overwrites an earlier one.
A verify call re-hashes the stored payload and reports whether it still matches.
What you get.
- ✓Frozen risk-report snapshot with a content hash
- ✓Verification that re-hashes and compares
- ✓No edit and no delete — by design, not by omission
- ✓Repeat freezes allowed, so a later pack can prove an earlier one is unchanged
- ✓Sits alongside the append-only audit log, which is enforced by database triggers
Quick answers.
Because a record you can delete is not evidence. If an analysis was wrong, freeze the corrected one — both will be there.
No. It proves what the analysis said and that it has not been altered since. Those are different claims.
More in AI Controls & Guardrails.
Every citation comes back with a verdict — verified, ambiguous or unresolved — and a match on title alone is never verified.
Each legal claim we hold about a jurisdiction is checked against that government's own legislation service — and where it cannot be, the answer is "unverified", not silence.
We do not train models on customer content. There is no opt-in, because there is no training programme.
User-pasted text is sanitised before reaching the LLM.