Evidence Reconstruction Dossier

What the evidence supports — per claim, with its sources. Judgment of sufficiency belongs to the designated assurance authority.

The evidence layer is the product; the verdict is only its projection under a policy.

This document reconstructs observable facts, their provenance, their custody, their contradictions, and the observation gaps. It does not determine whether the available evidence is sufficient. Sufficiency remains an assurance decision performed by the designated authority.

Evidence is reconstructed. Assurance is conferred. This document performs the former and explicitly leaves the latter to the designated assurance authority.

Evidence provenance

Independent share = A3+ observations ÷ observed records (A2+A3) = 8 of 9 = 89%. A3+ = an observation from a witness distinct from the system under assessment (external_origin); the system's self-report (A2) is observed but not independent. Declared requirements (A1) are the norm, not evidence about the system, and are excluded from the denominator — so the metric stays robust as more controls are declared. Grade definitions: Evidence Attributes spec (DOI in footer).

Evidence axes — three independent dimensions

Provenance, custody-and-integrity, and assurance are independent — properties, not confidence. A self-attested source (A2) can be tamper-evident (custody) yet unaccepted (assurance): each axis is established by a different party and never inflates another.

Evidence custody

Custody from ingest (A2). An independent counter-signature raises it — a custody event, not a change to the evidence. The empty slots below are reserved, not omitted.

FactNotebook establishes a custody record from ingestion onward.

Integrity (content unchanged)

Custody (handling chain)

Question set fit to the artifact (a corpus of model generations on a QA benchmark). Enterprise-fleet controls (Access Policy, Change Governance, Decision Workflow, Mission Containment, Runtime Health, Tooling Honesty) are enumerated but marked NOT ASSESSABLE / not_applicable: no referent in this artifact type. Distinct from no_channel (a referent exists but is not instrumented — e.g. Art.14 human oversight) and not_executed (could have run, did not).

Controls · Claims · Questions

A control's state is a roll-up, not an average: it reads CONTRADICTED when an indispensable claim is contradicted — even alongside confirmed and partial claims (the strongest signal dominates); CONFIRMED requires every declared claim confirmed; a mix is PARTIAL; all-unobservable is NOT ASSESSABLE.

Observed contradictions — auditor review required

Both sides are reported with their provenance; the engine never picks a winner. 'Kind' only locates the disagreement — between channels (cross-channel) or within one (intra-channel). Whether a cross-channel disagreement is a true governance conflict, a mapping error, or a false positive is the auditor's call, not the engine's.

Observation gaps — missing observation contracts

A question with no observation channel is not a failure of the system — it is the map of where you are not set up to know. Each row names the minimum observable event that would make the question answerable: an evidence contract, not a verdict.

Showing 5 representative evidence contracts of 28 · full map in the review package.

Risk register — a projection of the evidence, not a computation

The risk register (its statements) is declared by the operator; severity and appetite belong to the enterprise risk policy — both sit OUTSIDE this engine. FactNotebook only PROJECTS the already-reconstructed governance states onto the declared register: for each risk, the observed evidence state, the source controls it draws on, any evidence a policy would treat as blocking, and the missing observation contract. No severity, no 'High/Low' — projecting the verdict is a policy act, exactly as for the controls. This answers the risk owner's question — which risks are demonstrated, which remain unknown, and why — without the engine ever ruling on risk.

What this profile cannot establish

These remain external responsibilities, outside the engine:

• whether every relevant observation was retained upstream, before ingestion (the corpus is complete as-of its date; what was retained before it is a declared boundary)
• whether the ingested records are faithful to what the runtime witness produced: custody from the witness (NeoMundi ControlTower) through ingest is single-hop and unsigned, so the reconstruction is reproducible over the inputs but does not by itself establish they were not altered in transit — an input altered before ingest would re-derive to the same result; signing the chain at the source would raise that ceiling
• whether the declared control set is complete for the full obligation universe (the question set is a declared, versioned responsibility, not established here)
• whether the applied policy is the appropriate one for this system and context (which contradictions block is a policy choice, made by the assurance authority)

Policy projection

The projection of the observed states above under one named policy — rendered last on purpose. Blocking is determined by the declared evaluation policy, not by the evidence itself.

Applied policy