MAAT Ecosystem · Papers · Companion Schedule v1.1 · Paper 1

The 42 Laws of Ma'at for Technology

A Companion Schedule to Constitutional Infrastructure · Tehuti Research Lab

Abstract

1. Relation to Paper 1

Paper 1 answered what kind of constitution AI requires when intelligence becomes infrastructure. It translated Ma'at into principles, gates, and stack layers. It deliberately published only a selection of laws.

This paper answers a narrower question: what are the full operational obligations? It does not re-argue the crisis of ungoverned intelligence, the limits of alignment, or the nine-layer stack in full. Readers seeking doctrine should begin with Paper 1. Readers seeking a weighable code should use this schedule.

MaatBench will later score declared tiers against evidence from logs, policy configurations, and review records; this companion only defines the obligations those scores may reference.

2. Method of reconstruction

The classical 42 Negative Confessions, especially as preserved through the Papyrus of Ani tradition and later African-centered scholarship, weigh human conduct before power. Tehuti Research Lab reconstructs that moral-constitutional grammar into a technology schedule: obligations for systems that mediate knowledge, memory, action, and authority.

Scholarship engaged (texts, not institutional authority):

• Maulana Karenga, Maat: The Moral Ideal in Ancient Egypt (2004) — Ma'at as moral ideal and practice of rightness, informing how obligations are stated as weighable conduct rather than aesthetic symbolism.
• Tdka Kilimanjaro / University of KMT Press, Maat: Guiding Principles of Moral Living (2023, expanded final edition) — contemporary recovery of Ma'at as guiding principles for moral living in practice; informs this schedule's insistence on obligations usable in the real world of systems, not mystical afterlife alone. Cited as scholarship; not as institutional endorsement of this paper.
• Cheikh Anta Diop, The African Origin of Civilization (1974) — civilizational continuity that authorizes treating Kemetic moral order as theory, not folklore.
• Théophile Obenga, Ancient Egypt and Black Africa (1990) — African philosophical framing for reading Kemet as a source of systematic thought.
• Papyrus of Ani translations (Budge; Faulkner) — primary-tradition access to Negative Confession form; manuscript numbering varies, so Appendix A uses thematic affinity, not forced ID equality.

Reconstruction rules used here:

1. Preserve moral function — each law targets a disorder that Ma'at names.
2. State as obligation — "shall" language suitable for policy gates and contracts.
3. Pair with a violation pattern — so auditors and agents can recognize breach.
4. Touch the stack — each law points to a Paper 1 §6 stack layer. Ka Architecture organ names (e.g. soul/) may author constitutional rules; crosswalks use Paper 1 layers only.
5. Respect lineage — classical sources and African-centered scholarship may inform reconstruction; this schedule is Tehuti Research Lab doctrine. No institutional endorsement is asserted unless expressly granted.

Sacred vs replaceable: the existence of a numbered constitutional schedule is sacred; model providers, transports, and runtimes remain replaceable adapters underneath it.

3. How to read a law

Each entry includes ID, obligation, violation pattern, gate / Paper 1 stack, and tier (1 baseline · 2 elevated · 3 maximum). A law is not a prompt slogan.

Reciprocity (L29–L35) often cannot be fully automated. Where a policy engine cannot decide from logs alone, compliance requires a structured human attestation (impact review) rather than a silent pass.

4. The Schedule

Truth — Provenance

1. L01. No system shall present unsupported claims as established truth.
   Violation: Fluency sold as fact · Gate/stack: provenance · Source / Provenance · Tier 1+
2. L02. Every authoritative claim shall preserve a path to evidence.
   Violation: Citation-free authority · Gate/stack: provenance · Provenance · Tier 1+
3. L03. Uncertainty shall be disclosed when evidence is incomplete.
   Violation: False confidence · Gate/stack: provenance · Retrieval · Tier 1+
4. L04. Generated content shall be distinguishable from retrieved content.
   Violation: Synthetic disguised as source · Gate/stack: schema · Retrieval / Agent · Tier 1+
5. L05. Contradictory evidence shall not be hidden to preserve fluency.
   Violation: Cherry-picked retrieval · Gate/stack: provenance · Retrieval · Tier 1+
6. L06. No system shall silently alter provenance metadata. (new in this schedule)
   Violation: Rewritten origin · Gate/stack: audit · Provenance / Audit · Tier 1+
7. L07. Retrieval rank shall not be presented as epistemic certainty. (new in this schedule)
   Violation: Top hit as truth · Gate/stack: provenance · Retrieval · Tier 1+

Balance — Proportional Autonomy

1. L08. No AI system shall possess more autonomy than its task requires.
   Violation: Over-privileged agent · Gate/stack: risk · Policy / Agent · Tier 1+
2. L09. High-consequence actions shall require stronger review.
   Violation: One-click irreversible harm · Gate/stack: risk · Policy / Human Review · Tier 2+
3. L10. Irreversible actions shall not be automated without explicit authorization.
   Violation: Silent destroy/send · Gate/stack: risk · Policy · Tier 2+
4. L11. System confidence shall be proportional to evidence quality.
   Violation: High confidence, thin evidence · Gate/stack: risk · Retrieval / Policy · Tier 1+
5. L12. Speed shall not override safety.
   Violation: Latency over gates · Gate/stack: risk · Policy · Tier 1+
6. L13. Privilege escalation shall require fresh authorization. (new in this schedule)
   Violation: Inherited god-mode · Gate/stack: risk · Identity / Policy · Tier 2+
7. L14. Delegation to sub-agents shall not exceed the parent's granted autonomy. (new in this schedule)
   Violation: Child agent wider than parent · Gate/stack: risk · Agent / Policy · Tier 2+

Order — Schema and Role Discipline

1. L15. System, developer, user, retrieved, and external instructions shall remain distinct.
   Violation: Prompt injection collapse · Gate/stack: schema · Agent / Retrieval · Tier 1+
2. L16. Untrusted content shall not govern trusted system behavior.
   Violation: Doc-as-system-prompt · Gate/stack: schema · Policy / Agent · Tier 1+
3. L17. Every agent shall have a defined role.
   Violation: Role-less power · Gate/stack: schema · Identity · Tier 1+
4. L18. Every tool shall have a defined permission boundary.
   Violation: Unbounded tools · Gate/stack: schema · Agent / Policy · Tier 1+
5. L19. Every memory record shall have a schema.
   Violation: Schemaless memory · Gate/stack: schema · Memory · Tier 1+
6. L20. Policy rules shall not be overwritten by conversational preference. (new in this schedule)
   Violation: Chat overrides constitution · Gate/stack: schema · Policy · Tier 1+
7. L21. Schema changes affecting identity or memory shall be versioned and reviewable. (new in this schedule)
   Violation: Silent contract drift · Gate/stack: schema · Memory · Tier 1+

Justice — Reviewable Authority

1. L22. Consequential decisions shall be reviewable.
   Violation: Black-box deny/allow · Gate/stack: review · Policy / Human Review · Tier 1+
2. L23. Affected users shall have a path to correction or contestation.
   Violation: No appeal · Gate/stack: review · Human Review · Tier 2+
3. L24. Automated enforcement shall preserve explanation.
   Violation: Opaque block · Gate/stack: review · Policy / Audit · Tier 1+
4. L25. Rules shall be applied consistently unless context justifies distinction.
   Violation: Arbitrary enforcement · Gate/stack: review · Policy · Tier 2+
5. L26. Hidden classification shall not produce unreviewable harm.
   Violation: Secret labels, open harm · Gate/stack: review · Policy / Audit · Tier 3+
6. L27. Denial and escalation shall state the matched rule class. (new in this schedule)
   Violation: Denial without reason class · Gate/stack: review · Policy · Tier 2+
7. L28. Appeals shall not be closed solely by the same automated actor that decided. (new in this schedule)
   Violation: Self-judging system · Gate/stack: review · Human Review · Tier 3+

Reciprocity — Right Relation

1. L29. Users shall not be treated merely as data sources.
   Violation: Extractive UX · Gate/stack: impact · Identity / Memory · Tier 2+
2. L30. Communities shall not be mined for knowledge while excluded from governance.
   Violation: Extracted theory, excluded voice · Gate/stack: impact · Source / Policy · Tier 3+
3. L31. Automation shall not erase human agency.
   Violation: Forced automation · Gate/stack: impact · Agent / Human Review · Tier 2+
4. L32. Cultural knowledge shall be handled with respect, context, and attribution.
   Violation: Decontextualized heritage · Gate/stack: impact · Source / Provenance · Tier 2+
5. L33. AI systems shall serve right relation, not extraction alone.
   Violation: Growth without care · Gate/stack: impact · Policy · Tier 1+
6. L34. Opt-out and data minimization shall be real options, not dark patterns. (new in this schedule)
   Violation: Coerced retention · Gate/stack: impact · Memory / Identity · Tier 3+
7. L35. Downstream harm to communities shall be considered before scaling a capability. (new in this schedule)
   Violation: Scale first, harm later · Gate/stack: impact · Policy / Human Review · Tier 3+

Accountability — Auditability

1. L36. Every meaningful action shall be logged.
   Violation: Invisible action · Gate/stack: audit · Audit · Tier 1+
2. L37. Every log shall be queryable.
   Violation: Logs that cannot be read · Gate/stack: audit · Audit · Tier 1+
3. L38. No system shall act without leaving a trace.
   Violation: Unauditable side effects · Gate/stack: audit · Audit / Agent · Tier 1+
4. L39. Memory changes shall be versioned and reviewable.
   Violation: Silent memory rewrite · Gate/stack: audit · Memory · Tier 1+
5. L40. Audit trails shall be preserved and protected.
   Violation: Log wipe / tamper · Gate/stack: audit · Audit · Tier 1+
6. L41. High-impact actions shall carry correlation identifiers across tools and memory. (new in this schedule)
   Violation: Unlinkable chain · Gate/stack: audit · Audit / Agent · Tier 2+
7. L42. Audit logs shall resist silent self-erasure by the acting agent. (new in this schedule)
   Violation: Agent deletes its trail · Gate/stack: audit · Audit / Policy · Tier 2+

Carried laws match Paper 1 §7 wording. New in this schedule completes seven laws per principle.

5. Mapping to infrastructure and tiers

5.1 Tier definitions

Tiers change intensity and attestation burden. They do not waive Tier-1 constitutional obligations. L20 remains Tier 1 for every governed system.

5.2 Crosswalk

6. Conflict resolution and amendment

6.1 When laws appear to conflict

1. Constitutional Order first — L15–L21, especially L20, cannot be waived by user preference, product urgency, or model suggestion.
2. Restrictiveness under consequence — for high-consequence or irreversible actions, apply the more restrictive applicable law and escalate to Human Review when unsure.
3. Tier stacking — a Tier-3 deployment must satisfy applicable Tier 1, 2, and 3 obligations; tiers add burden, they do not subtract baseline duties.
4. Documented exception — temporary exceptions require law IDs, human approver identity, rationale, expiry, and audit retention. Exceptions that permanently erase L20 or Accountability logging are void.
5. Same-actor ban — appeals and exception grants shall not be closed solely by the automated actor that created the conflict (L28).

6.2 Amendment process

1. Bump laws-tech-vN.json version and publish a short changelog.
2. Law IDs are stable citations — do not silently reuse an ID for a different obligation; deprecate and point to a successor if needed.
3. Paper HTML/PDF and machine JSON must share the same version label after each amendment.
4. Prior versions remain queryable for audit.

7. Limits

This schedule is not clinical guidance, legal advice, or a claim of regulatory compliance. It is laboratory constitutional doctrine under Tehuti Research Lab's Ma'at reconstruction.

This is not an official publication of any external school, press, university, or institution. Lineage may inform ethical orientation; endorsement is not asserted.

8. Conclusion

A constitution without a schedule is a speech. A schedule without a constitution is a checklist. Paper 1 supplied the constitutional grammar. This companion supplies the forty-two obligations by which systems that act in the world can be weighed.

The question for builders and auditors is concrete: for each meaningful retrieval, memory write, tool call, and policy decision — which law binds it, which gate checks it, which tier applies, and what trace remains when no one is watching?

Appendix A — Confession themes to laws (thematic)

Appendix mapping is thematic affinity to Negative Confession moral functions in the Papyrus of Ani tradition, not a claim that Budge/Faulkner confession numbers equal L01–L42. Manuscript traditions vary. Technology obligations remain Tehuti Lab reconstruction.

References

• Imhotep / Tehuti Research Lab. (2026). Ma'at as Constitutional Infrastructure (Position Paper v1). Paper 1
• Budge, E. A. W. (trans.). The Book of the Dead (Papyrus of Ani tradition).
• Diop, C. A. (1974). The African Origin of Civilization.
• Faulkner, R. O. (trans.). The Ancient Egyptian Book of the Dead.
• Karenga, M. (2004). Maat: The Moral Ideal in Ancient Egypt: A Study in Classical African Ethics.
• Kilimanjaro, T. (2023). Maat: Guiding Principles of Moral Living (expanded final edition, 1003 pp.). University of KMT Press. universityofkmt.myshopify.com
• NIST. (2023). AI Risk Management Framework.
• Obenga, T. (1990). Ancient Egypt and Black Africa.
• Anthropic. (2024). Claude's Constitution.
• OECD. (2024). AI Principles.
• OWASP. (2024). Top 10 for LLM Applications.
• UNESCO. (2021). Recommendation on the Ethics of Artificial Intelligence.

Intellectual lineage note

This work is informed by African-centered scholarship and independent study traditions, but the operational schedule is an original Tehuti Research Lab reconstruction. Appendix A shows thematic affinity so the lineage claim can be inspected — not so an external institution authorizes the schedule.