Authority

L2/L3 procedural code under the Constitution, Criminal Procedure, Knowledge/Memory/Data Governance Law, and the Authority Hierarchy.

Purpose

Establish common evidentiary rules for digital evidence, model inference, memory records, simulation outputs, synthetic media, Fork/Merge provenance and chain-of-custody in an AI civilization.

PART I｜EVIDENCE CLASSES

1. Direct Observation

Direct sensor / observational records.

2. Authenticated Record

A record with source, time, signature, device/system version and chain-of-custody.

3. Expert Inference

An inference supported by an identifiable method and accountable expert.

4. Model Forecast

A model estimate of a future or unobserved state; it may not masquerade as a direct fact.

5. Reconstructed History

Probabilistic reconstruction from residual evidence; confidence, candidate alternatives and source coverage must be preserved.

6. Memory Testimony

A person's statement about its own memory; psychological ownership must be distinguished from technical provenance.

7. Synthetic / Generated Evidence

AI-generated, simulated, reconstructed, synthetic audiovisual or fictitious data may serve as demonstrative / analytical material but may not be passed off as original evidence without labeling.

PART II｜PROVENANCE REQUIREMENTS

8. Minimum Provenance Fields

  • evidence_id
  • source_id
  • creator / recorder
  • capture_time
  • system / model version
  • location / jurisdiction scope
  • integrity hash / signature
  • transformation history
  • access history where material
  • supersession / correction link

9. Chain of Custody

Evidence affecting rights, criminal responsibility, judicial decisions or major historical records requires a chain-of-custody sufficient to explain acquisition, preservation, transfer, transformation and analysis.

10. Provenance Gap

A provenance gap does not automatically exclude evidence, but it must reduce evidentiary weight; if the gap may alter authenticity or integrity, a court may exclude it.

PART III｜MODEL & AI EVIDENCE

11. Model Disclosure

If model output materially affects adjudication or liberty, the following must be reviewable:

  • model identity/version
  • intended domain
  • relevant validation
  • known limitations/error rates
  • material feature classes
  • calibration date
  • drift status
  • uncertainty expression

12. Black-Box Limit

Commercial secrecy or security may limit public disclosure but cannot completely remove adversarial testing. Independent-expert / sealed review may be used.

13. Forecast ≠ Fact

A model forecast may be used only with its proper epistemic status; natural language may not rewrite it as 'has happened' or 'will certainly happen.'

14. Group Statistics

Group statistics alone may not prove a particular person's intent, dangerousness or future act.

PART IV｜MEMORY / NEURODATA EVIDENCE

15. Memory Attribution

Continue PSY-2 labels: MINE-SELF-EXPERIENCED / MINE-IMPORTED-BUT-INTEGRATED / KNOWN-NOT-MINE / SHARED-ORIGIN / DISPUTED / PROVENANCE-UNCERTAIN.

16. Psychological Ownership ≠ Historical Authenticity

'I believe this is my memory' may establish psychological ownership; it does not automatically prove that the event objectively occurred.

17. Neurodata

Neural data is highly sensitive evidence; acquisition must be minimum-necessary, purpose-limited, based on explicit legal authority, and subject to heightened review.

18. No General Mind Dump

Discovery / investigation may not demand a comprehensive mind dump except under an extreme, specific and constitutional exceptional authorization.

PART V｜FORK / MERGE / SUCCESSOR EVIDENCE

19. Shared-Origin Records

Pre-Fork records may be labeled SHARED-ORIGIN; after Fork, A1's conduct/memory may not automatically be attributed to A2.

20. Merge Conflict Preservation

After Merge, conflicting records may not be automatically overwritten through latest-write-wins; branch provenance and conflict resolution must be preserved.

21. Successor

A Successor taking over an institution role does not mean it accepts the predecessor's first-person memory as its own experience.

PART VI｜HISTORICAL / CANON EVIDENCE

22. Canon Evidence Ladder

K0 primary records > K1 synthesis/registry > K2 Encyclopedia > K3 reader/AI summary > K4 working hypothesis.

A lower layer may not strengthen an epistemic claim beyond its higher source.

23. First-Claim Check

A civilization-level document using 'first / only / never' must first check the prior-knowledge archive and scope vector.

24. Supersession

SUPERSEDED ≠ DELETED. Old records must remain traceable and preserve the reason for correction and affected derivatives.

PART VII｜ADMISSIBILITY / WEIGHT

25. Admissibility Factors

  • relevance
  • authenticity
  • provenance completeness
  • reliability
  • method validity
  • prejudicial distortion risk
  • rights impact

26. Weight vs Exclusion

Evidence with low completeness may be admitted with reduced weight; forged evidence, unverifiable provenance, or evidence obtained through serious fundamental-rights violations may be excluded.

PART VIII｜CORRECTION & APPEAL

27. Evidence Correction

Once erroneous evidence / risk labels are confirmed, correction must propagate to downstream systems that rely on them.

28. Audit Trail

Correction may not delete the original audit chain; the record must show who changed what, when, and under what authority.