# The AI Evidence Audit Checklist

Know what to trust before you cite it.

## Product Promise
A structured, field-tested checklist that helps researchers and knowledge workers classify every AI-generated claim as verified, inferred, or hallucinated before it enters a document, report, or decision. Validated in the LegiPro Mirofish pilot across 34 dossiers and 68 turns.

## Intended Buyer
Researchers, knowledge workers, consultants, policy teams, and any professional using AI in evidence-critical writing who needs a defensible classification trail before output enters a document or decision.

## Included Deliverables
- 50-item audit checklist (PDF + Notion template)
- 4-tier classification rubric: Verified / Inferred / Plausible / Hallucinated
- Quick-reference card (1-page printable)
- 5 worked examples across disciplines
- LegiPro pilot evidence notes

## Practitioner Workflow
1. Source verification: define the decision, evidence, owner, and acceptance threshold before use.
2. Claim classification: define the decision, evidence, owner, and acceptance threshold before use.
3. Logical soundness: define the decision, evidence, owner, and acceptance threshold before use.
4. Epistemic risk: define the decision, evidence, owner, and acceptance threshold before use.
5. Documentation and traceability: define the decision, evidence, owner, and acceptance threshold before use.

## Operating Standard
Use this product as a practical governance aid, not as legal advice. For legal, regulatory, medical, financial, or employment decisions, require qualified human review and preserve the evidence trail.

## Evidence Rules
- Prefer primary sources over AI-generated summaries.
- Keep the raw AI output, prompts, model name, date, and reviewer identity.
- Separate verified claims from inferred, plausible, and unsupported claims.
- Do not cite AI output as a substitute for a real source.
- For volatile law, policy, science, or market claims, re-check the source close to publication.

## Official Source Anchors
- Quebec Act respecting the protection of personal information in the private sector, CQLR c P-39.1: https://www.legisquebec.gouv.qc.ca/en/document/cs/p-39.1/20240701
- NIST AI Risk Management Framework 1.0: https://www.nist.gov/itl/ai-risk-management-framework
- ISO/IEC 42001:2023 AI management systems: https://www.iso.org/standard/42001