Decoupling Attribution from Inference: Strengthening Legal-Geometric Auditing (RFC-006) via Deterministic Gates, Cryptographic Seams, and Anti-Conflation Provenance (RFC-007)

📌 TECHNICAL DISCLOSURE & FOLLOW-UP PREPRINT ARCHIVE (RFC-007) This record archives Request for Comments (RFC-007): "Decoupling Attribution from Inference: Strengthening Legal-Geometric Auditing (RFC-006) via Deterministic Gates, Cryptographic Seams, and Anti-Conflation Provenance." It serves as a methodological audit and architectural upgrade to RFC-006 ("The Geometry of Institutional Enclosure", DOI: https://doi.org). Related Source Repositories & Artifacts: • Legacy Preprint (RFC-006): https://doi.org • deterministic-ai-gate (v1.0, MIT): https://github.com/Borealiscodes/deterministic-ai-gate • active-inference-commit-seam (v0.1.0, Dual-Boundary MIT / CAUSA NC 1.1): https://github.com/Borealiscodes/active-inference-commit-seam/tree/main • Upstream CAUSA Library (Stell, CAUSA Non-Commercial License 1.1): https://github.com/stell2026/Causa ⚠️ EXPLICIT ATTRIBUTION CLARIFICATION & EPISTEMIC STATEMENT To maintain strict scholarly and engineering provenance, this record issues three explicit clarifications regarding RFC-006 and upstream software dependencies: 1. Zero Normative Attribution to Stell: Stell—author of the Anima cognitive architecture and the stell2026/Causa causal agency library—has NEVER made any statements, claims, or legal arguments regarding transnational repression, administrative denial of remedy, or international human-rights treaties. CAUSA provides domain-agnostic mathematical metrics (directional ownership and expression congruence) over abstract numeric vectors. Any application of those metrics to legal or administrative case files rests solely with the downstream researcher (NDH-RESEARCH-PILOT). 2. Empirical Audit of LLM Citation & Tool Conflation: The blurred attribution layers present in the legacy RFC-006 draft are analyzed here as an instructive, real-world demonstration of how Large Language Models (LLMs) conflate co-occurring context windows—attaching normative legal statements to technical libraries and inventing unified annotation narratives. This meta-finding directly motivates replacing LLM-managed citations with fail-closed deterministic CPU verification gates. 3. Reference Prototype & Dual-License Boundary: The companion software baselines (deterministic-ai-gate and active-inference-commit-seam) are clean-room reference prototypes requiring domain-specific empirical validation prior to production deployment. Core verification modules are MIT-licensed with zero external dependencies, while the optional CAUSA telemetry adapter is isolated behind an explicit boundary respecting the CAUSA Non-Commercial License 1.1. 🛡️ SUMMARY OF CONTRIBUTIONS • Clean-Room Input/Output Verification (deterministic-ai-gate): Locks raw legal and administrative source documents into pre-LLM SHA-256 chunk manifests, constrains LLMs strictly to schema-bounded text-to-JSON parsing, and verifies quotes, 2048-bit bounded rational arithmetic (Q_B), and Kahn topological DAG timelines deterministically on CPU. • Proposal-to-Commit Architectural Seam (active-inference-commit-seam): Isolates exploratory triage (Salience Δ, Structural Resonance Φ, Active Inference queries, and optional CAUSA permutation baselines) on the proposal side of a strict commit boundary (Record ≠ Admit ≠ Commit), enforcing the Law of Non-Self-Warranting Memory (repetition ≠ evidence). • Formal Separation of External Warrant & Internal Agency: Defines admissibility into evidentiary ledgers as the strict conjunction of external SHA-256 provenance (W_ext) and statistically significant internal agency above shuffled permutation baselines (A_int), ensuring internal vector coherence can never substitute for primary source truth.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24275846
Primary Topic
Ethics and Social Impacts of AI
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Decoupling Attribution from Inference: Strengthening Legal-Geometric Auditing (RFC-006) via Deterministic Gates, Cryptographic Seams, and Anti-Conflation Provenance (RFC-007)

Hedling Borealis
5 links
Zenodo (CERN European Organization for Nuclear Research)
Ethics and Social Impacts of AI
preprint

Decoupling Attribution from Inference: Strengthening Legal-Geometric Auditing (RFC-006) via Deterministic Gates, Cryptographic Seams, and Anti-Conflation Provenance (RFC-007)

Hedling Borealis
preprint en

Abstract

📌 TECHNICAL DISCLOSURE & FOLLOW-UP PREPRINT ARCHIVE (RFC-007) This record archives Request for Comments (RFC-007): "Decoupling Attribution from Inference: Strengthening Legal-Geometric Auditing (RFC-006) via Deterministic Gates, Cryptographic Seams, and Anti-Conflation Provenance." It serves as a methodological audit and architectural upgrade to RFC-006 ("The Geometry of Institutional Enclosure", DOI: https://doi.org). Related Source Repositories & Artifacts: • Legacy Preprint (RFC-006): https://doi.org • deterministic-ai-gate (v1.0, MIT): https://github.com/Borealiscodes/deterministic-ai-gate • active-inference-commit-seam (v0.1.0, Dual-Boundary MIT / CAUSA NC 1.1): https://github.com/Borealiscodes/active-inference-commit-seam/tree/main • Upstream CAUSA Library (Stell, CAUSA Non-Commercial License 1.1): https://github.com/stell2026/Causa ⚠️ EXPLICIT ATTRIBUTION CLARIFICATION & EPISTEMIC STATEMENT To maintain strict scholarly and engineering provenance, this record issues three explicit clarifications regarding RFC-006 and upstream software dependencies: 1. Zero Normative Attribution to Stell: Stell—author of the Anima cognitive architecture and the stell2026/Causa causal agency library—has NEVER made any statements, claims, or legal arguments regarding transnational repression, administrative denial of remedy, or international human-rights treaties. CAUSA provides domain-agnostic mathematical metrics (directional ownership and expression congruence) over abstract numeric vectors. Any application of those metrics to legal or administrative case files rests solely with the downstream researcher (NDH-RESEARCH-PILOT). 2. Empirical Audit of LLM Citation & Tool Conflation: The blurred attribution layers present in the legacy RFC-006 draft are analyzed here as an instructive, real-world demonstration of how Large Language Models (LLMs) conflate co-occurring context windows—attaching normative legal statements to technical libraries and inventing unified annotation narratives. This meta-finding directly motivates replacing LLM-managed citations with fail-closed deterministic CPU verification gates. 3. Reference Prototype & Dual-License Boundary: The companion software baselines (deterministic-ai-gate and active-inference-commit-seam) are clean-room reference prototypes requiring domain-specific empirical validation prior to production deployment. Core verification modules are MIT-licensed with zero external dependencies, while the optional CAUSA telemetry adapter is isolated behind an explicit boundary respecting the CAUSA Non-Commercial License 1.1. 🛡️ SUMMARY OF CONTRIBUTIONS • Clean-Room Input/Output Verification (deterministic-ai-gate): Locks raw legal and administrative source documents into pre-LLM SHA-256 chunk manifests, constrains LLMs strictly to schema-bounded text-to-JSON parsing, and verifies quotes, 2048-bit bounded rational arithmetic (Q_B), and Kahn topological DAG timelines deterministically on CPU. • Proposal-to-Commit Architectural Seam (active-inference-commit-seam): Isolates exploratory triage (Salience Δ, Structural Resonance Φ, Active Inference queries, and optional CAUSA permutation baselines) on the proposal side of a strict commit boundary (Record ≠ Admit ≠ Commit), enforcing the Law of Non-Self-Warranting Memory (repetition ≠ evidence). • Formal Separation of External Warrant & Internal Agency: Defines admissibility into evidentiary ledgers as the strict conjunction of external SHA-256 provenance (W_ext) and statistically significant internal agency above shuffled permutation baselines (A_int), ensuring internal vector coherence can never substitute for primary source truth.

Zenodo (CERN European Organization for Nuclear Research)
Peace, justice, and strong institutions
Ethics and Social Impacts of AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon