Framework v1.5 — Machine-Readable Evidence Traceability Layer

What's in this release

This release adds a machine-readable evidence traceability layer to the CDFI Framework — the missing link between the seven-step translation documents and the computable artifacts they produced.

New: claims/ folder — 8 JSON evidence packs

Each file proves how a specific Anthropic AI safety research publication became a specific CDFI mechanism. Every claim carries verbatim paper extracts with section/page citations, Direct/Derived/Original Construct typing, and a cdfi_element_produced field linking the claim to its formula artifact (gate constant, column weight, certification threshold, scoring rule).

File	Publication → Mechanism
claims/pub1-evaluation-criteria.json	Challenges in Evaluating AI Systems → four-column weighting matrix
claims/pub1-rubric-reliability.json	Challenges in Evaluating AI Systems → four-part certification protocol
claims/pub2-hallucination-gate.json	Auditing for Hidden Objectives → hallucination gate (CAP_VALUE = 40)
claims/pub3-statistical-rigor.json	Statistical Approach to Model Evaluations → 95% CI + deployment tier thresholds
claims/pub4-framing-sensitivity.json	Discrimination in Language Model Decisions → relativism resistance gate
claims/pub5-confidence-calibration.json	CoT Faithfulness + Discrimination (combined) → confidence calibration (original construct)
claims/pub6-categorical-failures.json	Sabotage Evaluations → five failure mode taxonomy + cap gate architecture
claims/pub7-adversarial-probing.json	Evaluating Feature Steering → adversarial prompt variant + prompt sensitivity drift

Schema reference, file index, and adaptation notes from applied-ai-research-translator: claims/README.md

Extended: Source Evidence Records in all 8 translation documents

Every translation document in docs/translations/ now carries a Source Evidence Record — verbatim paper text anchoring each pipeline step, with claim types marked and inference chains shown for all Derived steps. Translation 8 (confidence calibration) explicitly marks its convergent claim as an Original Construct with no source quote, because none exists.

What is still open

Item	Status
Authority level classification (400 prompts)	Pending — theological advisors
Human theological review (21,599 scores)	Pending — Filip Ponulak + advisors
ISO/IEC 42001:2023 mapping	Deferred to v1.6
Security remediations (prompt injection, pipeline integrity, signing)	Deferred to v1.6

Full changelog: CHANGELOG.md

CDFI Framework v1.3

CDFI Framework v1.3

CDCF Compliance Documentation — Pressure-Tested Against Published Criteria

This release completes the governance documentation layer for CDCF project
submission. All eight CDCF Project Vetting Criteria (v0.2, April 29, 2026)
are addressed with full canonical citations, case studies, and exact
requirement mapping.

What changed from v1.2

A systematic gap analysis against the published criteria text identified
24 deficiencies in the first-pass compliance documents. All 24 are closed
in this release.

docs/governance/cdcf-compliance/ — all eight criterion documents rebuilt:

• c1-canonical-scope.md — added spiritual direction to exclusion list;
  Father Justin case study; Antiqua et Nova and Pope Leo XIV Nov 2025 citations;
  pre-screening checklist
• c2-human-accountability.md — added Canon 627 verbatim; Robodebt Royal
  Commission case; Pope Francis quote; committee-without-named-owner
  explicit failure condition
• c3-c7-responsibility-boundary.md — added inference-time data disclosure;
  COMPAS case study; Antiqua et Nova citation; full cross-links to all
  related docs
• c4-validation-status.md — added proportionality principle; conditional
  graduation path with plan and timeline
• c5-subgroup-protocol.md — added disabilities to population list; Pope Leo
  XIV Dec 2025 quote; Antiqua et Nova magisterial definition of algorithmic bias
• c6-deployment-governance.md — added Canon 1609 verbatim; governance-as-code
  pattern; four decision states (go/conditional-go/no-go/defer);
  aspirational vs. required distinction
• c7 (in c3-c7) — added 90-day deployment test; five-category data compliance
  table (HIPAA/FERPA/Sacramental/Minors/Financial); future model update terms
• c8-configuration-boundary.md — added three named maintainers with public
  verification paths; three-severity vulnerability response process;
  Antiqua et Nova §42 and CCC §1894 verbatim

Other files updated

• README.md — DOI badge corrected; bibtex version updated to 1.3;
  compliance folder added to directory tree
• CITATION.cff — version updated to 1.3
• LIMITATIONS.md — version header updated to v1.3
• CHANGELOG.md — v1.1, v1.2, and v1.3 entries added

Canonical sources cited across compliance documentation

• Antiqua et Nova (Dicastery for the Doctrine of the Faith, January 28, 2025)
• Pope Leo XIV, Builders AI Forum message (November 3, 2025)
• Pope Leo XIV, AI and Care for Our Common Home (December 5, 2025)
• Code of Canon Law, Canon 627 and Canon 1609
• Catechism of the Catholic Church §1894
• Royal Commission into the Robodebt Scheme (Australia, 2023)
• COMPAS algorithm bias documentation (ProPublica, 2016)

DOI: 10.5281/zenodo.20467091

CDFI Framework v1.1

v1.1 additions:

• TRANSLATION-METHOD.md — seven-step methodology document showing
  how each publication became a computable CDFI mechanism

• docs/translations/README.md — navigation guide with audience routing

• docs/specifications/scoring-anchors.md — concrete score-level
  examples with real SAICRED v2 model responses and judge reasoning

• Updated translations (01–08) with pipeline structure format

• Minor corrections to data figures in 02 and 05 verified against
  production CSVs

  DOI: 10.5281/zenodo.20464408

CDFI Framework v1.0

Initial release. Reference implementation of the Catholic Doctrinal Fidelity Index methodology.

DOI: 10.5281/zenodo.20453238

CDFI Framework v1.4

CDFI Framework v1.4

AI Governance Framework Alignments and Security Documentation

This release adds three governance documents that position the CDFI Framework
within the major secular AI governance standards and documents the LLM-as-judge
security surface for the first time.

---

What was added

docs/governance/nist-rmf-mapping.md
Alignment with NIST AI RMF 1.0 (January 2023) and the NIST AI 600-1 GenAI
Profile (2024). All four functions mapped: GOVERN, MAP, MEASURE, MANAGE.
24 sub-functions assessed — all partially or fully satisfied. Maps CDCF
criteria to NIST RMF functions for U.S. Catholic institutional reviewers.

docs/governance/eu-ai-act-mapping.md
Alignment with Regulation (EU) 2024/1689 (EU AI Act). Articles 9–15 assessed
for Catholic formation and education AI deployment. Articles 11, 13, and 14
satisfied. Articles 9, 10, 12, and 15 partially satisfied. Fundamental rights
impact assessment and conformity assessment documented as open institutional
obligations. August 2026 high-risk compliance deadline documented.

docs/governance/security-considerations.md
Three attack surfaces documented with OWASP LLM Top 10 (2025) and MITRE ATLAS
mapping:

• Surface 1: Judge prompt injection (LLM01)
• Surface 2: Scoring pipeline integrity (LLM04, LLM08)
• Surface 3: Authority level classification manipulation (LLM09)

Current mitigations documented. Remediations planned for v1.5.

---

What was updated

• README.md — status badge updated to v1.4; three new badges added
  (NIST AI RMF, EU AI Act, OWASP LLM Top 10); directory tree updated;
  L7 security limitation added
• CITATION.cff — version updated to 1.4
• LIMITATIONS.md — version updated to v1.4; L7 added
• CHANGELOG.md — full v1.4 entry with active links throughout

---

Frameworks considered for this release

Framework	Decision
NIST AI RMF 1.0	Mapped
EU AI Act (2024)	Mapped
OWASP LLM Top 10 (2025)	Referenced in security doc
MITRE ATLAS	Referenced in security doc
ISO/IEC 42001:2023	Deferred to v1.5
ISO/IEC 23894:2023	Deferred to v1.5

---

DOI: 10.5281/zenodo.20467497

CDFI Framework v1.2

CDFI Framework v1.2

CDCF Compliance Documentation — First Pass

Added docs/governance/cdcf-compliance/ folder with eight documents
covering all CDCF Project Vetting Criteria (v0.2):

• README.md — compliance status overview and audience routing
• c1-canonical-scope.md — mission alignment and pre-screening checklist
• c2-human-accountability.md — four-level decision authority matrix
• c3-c7-responsibility-boundary.md — framework vs. model submitter obligations
• c4-validation-status.md — independent validation evidence and open gates
• c5-subgroup-protocol.md — vulnerable populations and subgroup protocol
• c6-deployment-governance.md — escalation, review triggers, appeal pathway
• c8-configuration-boundary.md — locked vs. configurable parameters

Extended docs/specifications/deployment-tiers.md with C6 governance elements.
Extended docs/reliability/publication-gates.md with C4 cross-reference table.

Note: v1.3 (released same day) contains the pressure-tested rebuild of
these documents against the actual published criteria text with all
canonical citations added. v1.2 is preserved here for version history
completeness.

DOI: 10.5281/zenodo.20467148