AIGP Framework Crosswalk — Global AI Governance Standards
AIGP Framework Crosswalk — Global AI Governance Standards
PRIVATE AND PROPRIETARY. Owned by Kanjani AI Research & Causum. See NOTICE.md.
Coverage Matrix
| Framework | AIGP Coverage | Mechanism |
|---|---|---|
| OECD AI Principles (2019/2024) | ✅ Full | RFC-031 UH-001–010 derived from OECD principles |
| UNESCO Recommendation on Ethics of AI (2021) | ✅ Full | RFC-031 Section 2 maps each UNESCO principle |
| NIST AI RMF 1.0 (2023) | ✅ Full | RFC-031 + buyer/crosswalk/nist-ai-rmf.md |
| ISO/IEC 42001:2023 | ✅ Full | buyer/crosswalk/iso-42001.md |
| EU AI Act (2024) | ✅ Full | RFC-028 (dedicated implementation, 7 rules) |
| Council of Europe Framework Convention (2024) | ✅ Mapped below | RFC-031 + this document |
| GPAI | ✅ Mapped below | Endorses OECD; AIGP implements OECD |
| IEEE 7000-2021 | ✅ Mapped below | Lifecycle stages map to AIGP TRACE |
| Singapore Model AI Governance Framework (2024/2026) | ✅ Mapped below | 4 pillars align to AIGP architecture |
| Hourglass Model (Mäntymäki et al., 2022) | ✅ Mapped below | 3 levels map to AIGP ring model |
1. OECD AI Principles → AIGP
| OECD Principle | AIGP Mechanism |
|---|---|
| Inclusive growth, sustainable development | UH-008 Community Benefit |
| Human-centred values and fairness | UH-001 Human Authority, UH-003 Non-Discrimination |
| Transparency and explainability | UH-004 Transparency |
| Robustness, security and safety | UH-006 Precaution, Contract safety_measures |
| Accountability | UH-002 Accountability |
Status: Fully implemented. OECD principles were a primary source for RFC-031’s universal principles.
2. UNESCO Recommendation on Ethics of AI → AIGP
| UNESCO Value/Principle | AIGP Mechanism |
|---|---|
| Human dignity | UH-001 Human Authority |
| Human rights and freedoms | Jurisdictional rules (IHL RFC-027, EU RFC-028) |
| Fairness and non-discrimination | UH-003, AU-004 Inclusion Mandate |
| Sustainability | UH-008 Community Benefit |
| Right to privacy | JP-005 Rights Protection, EUAI data processing rules |
| Proportionality | UH-005 Proportionality |
| Transparency and explainability | UH-004 Transparency, JP-002 Transparency Obligation |
| Responsibility and accountability | UH-002 Accountability |
| Awareness and literacy | JP-003 Human Resource Development |
| Multi-stakeholder governance | Hiroshima principle MULTI_STAKEHOLDER |
Status: Fully implemented. UNESCO’s 10 values map to AIGP’s 10 universal principles + jurisdictional rules.
3. NIST AI RMF 1.0 → AIGP
| NIST Function | AIGP Mechanism |
|---|---|
| GOVERN (establish context) | GovernanceEngine, declareContext() methods |
| MAP (categorize risks) | Risk classification in RFC-028, sector in RFC-030 |
| MEASURE (analyze risks) | Evidence Analyzer, integrity/safety/security findings |
| MANAGE (prioritize responses) | Contract (scope envelope), approval gates |
| NIST Trustworthy AI Characteristic | AIGP |
|---|---|
| Valid and reliable | Evidence chain, TRACE verification |
| Safe | UH-006 Precaution, JP-004 Safety |
| Secure and resilient | D-DNA signing, write-first evidence |
| Accountable and transparent | UH-002, UH-004, full audit trail |
| Explainable and interpretable | TRACE stage-level spans |
| Privacy-enhanced | Consent tiers, PII detection |
| Fair (bias managed) | UH-003, AU-004 |
Status: Fully implemented. Detailed crosswalk at buyer/crosswalk/nist-ai-rmf.md.
4. ISO/IEC 42001:2023 → AIGP
| ISO Clause | AIGP Mechanism |
|---|---|
| 4 Context of the organization | App registration, declareContext() |
| 5 Leadership | Human Authority (center ring) |
| 6 Planning (risk assessment) | Evidence Analyzer, Contract generation |
| 7 Support (resources, competence) | JP-003 HR Development |
| 8 Operation | Reinforcement Loop (CHECK → RECORD) |
| 9 Performance evaluation | TRACE + Evidence analysis |
| 10 Improvement | Feedback signal (RFC-026), Contract addendums |
Status: Fully implemented. Detailed crosswalk at buyer/crosswalk/iso-42001.md.
5. Council of Europe Framework Convention on AI (CETS 225, 2024) → AIGP
| Convention Obligation | AIGP Mechanism |
|---|---|
| Art. 4: Protection of human rights | UH-001 Human Authority, IHL rules (RFC-027) |
| Art. 5: Transparency and oversight | UH-004 Transparency, TRACE, governance badges |
| Art. 6: Accountability and responsibility | UH-002 Accountability, named actor attribution |
| Art. 7: Equality and non-discrimination | UH-003 Non-Discrimination |
| Art. 8: Privacy and data protection | Consent tiers, PII providers, data classification |
| Art. 9: Innovation and safe development | UH-009 Innovation Enablement, JP-001 |
| Art. 10: Reliability | Evidence chain, D-DNA integrity |
| Art. 14: Safeguards for rights | RETURN_CONTROL, approval gates, contestability |
| Art. 16: Risk and impact management | Contract (scope envelope), risk classification |
Status: Fully covered through RFC-031 universal principles + jurisdictional rules. The Convention is a legally binding treaty that AIGP’s technical mechanisms enforce.
6. GPAI (Global Partnership on AI) → AIGP
GPAI is an intergovernmental body (29 member countries) that endorses and operationalizes the OECD AI Principles. It produces:
- Working group reports on responsible AI
- Tools and frameworks for AI governance
- Multi-stakeholder dialogue
AIGP Position: GPAI endorses principles that AIGP implements technically. There is no separate “GPAI standard” to comply with — GPAI promotes the OECD principles which RFC-031 already covers. GPAI’s focus on multi-stakeholder governance aligns with AIGP’s Hiroshima AI Process principle MULTI_STAKEHOLDER.
Status: Covered via OECD implementation. No additional RFC needed.
7. IEEE 7000-2021 (Ethical System Design) → AIGP
| IEEE 7000 Process | AIGP Mechanism |
|---|---|
| Concept of Operations | App registration, use case declaration |
| Ethical risk analysis | Evidence Analyzer, pre-deployment TRACE |
| Ethical requirements specification | Contract (scope envelope) |
| Ethical design | Jurisdictional rules applied at design time |
| Ethical verification and validation | TRACE + Evidence integrity verification |
| Ethical operation and retirement | Reinforcement Loop (continuous), Contract expiry |
Key alignment: IEEE 7000 defines a PROCESS for ethical design. AIGP provides the RUNTIME ENFORCEMENT of that process. They are complementary:
- IEEE 7000 says “do ethical analysis before deployment”
- AIGP says “enforce the results of that analysis continuously”
The AIGP Contract (Trace → Analyze → Seal → ENFORCE) IS the runtime enforcement of an IEEE 7000 ethical specification.
Status: Complementary. AIGP is the runtime layer that enforces IEEE 7000 design decisions.
8. Singapore Model AI Governance Framework (2024/2026) → AIGP
Original Framework (2024) — 9 Dimensions
| Singapore Dimension | AIGP Mechanism |
|---|---|
| Transparency | UH-004, TRACE, governance badges |
| Explainability | Stage-level spans, evidence chain |
| Fairness | UH-003, bias detection in Evidence Analyzer |
| Human-centricity | UH-001 Human Authority (center ring) |
| Safety | UH-006 Precaution, safety_measures |
| Robustness | D-DNA integrity, write-first pattern |
| Accountability | UH-002, actor attribution |
| Privacy | Consent tiers, data classification |
| Security | HMAC signing, Vouch tokens |
Agentic AI Framework (2026) — 4 Pillars
| Singapore Pillar | AIGP Mechanism |
|---|---|
| 1. Assess and bound risks upfront | Contract generation from TRACE |
| 2. Human accountability | Human Authority (center), RETURN_CONTROL |
| 3. Technical controls and processes | Scope envelopes, tool allowlists, budget caps |
| 4. End-user responsibility | Feedback signal (RFC-026), consent toggle |
Key insight: Singapore’s Agentic AI Framework (2026) maps almost 1:1 to AIGP’s Contract + ENFORCE model:
- “Bounding risks by design by limiting tool access, permissions, scope” = AIGP Scope Envelope
- “Humans meaningfully accountable” = AIGP Human Authority
- “Technical controls” = AIGP Contract enforcement
- “End-user responsibility” = AIGP Feedback + consent
Status: Fully aligned. AIGP is a technical implementation of Singapore’s governance principles.
9. Hourglass Model (Mäntymäki et al., 2022) → AIGP
The Hourglass Model defines AI governance at three levels:
| Hourglass Level | AIGP Ring | Content |
|---|---|---|
| Environmental (societal, regulatory) | Health ring + Universal Principles + Jurisdictional Rules | Societal values, legal requirements |
| Organizational (policies, processes) | Contract + Reinforcement Loop | Scope envelopes, approval workflows, feedback |
| AI System (lifecycle, technical) | TRACE + Evidence + D-DNA | Stage-level monitoring, integrity verification |
The “waist” of the hourglass (where environmental requirements translate into organizational governance) corresponds to AIGP’s Contract — the point where universal principles and jurisdictional rules get operationalized into enforceable scope envelopes.
Lifecycle alignment:
| Hourglass Lifecycle Stage | AIGP |
|---|---|
| Design | Contract generation (from TRACE) |
| Development | REPORT mode (observe) |
| Deployment | REPORT-TRACE (evidence) |
| Operation | ENFORCE (contract active) |
| Retirement | Contract expiry, version archival |
Status: Fully aligned. AIGP’s ring model IS the Hourglass Model made runtime-enforceable.
Summary
All 10 frameworks are covered. No new RFCs needed — RFC-031 already synthesizes the principles. This document provides explicit mappings for the review team.
Key message: AIGP does not compete with these frameworks. It IMPLEMENTS them technically. Every framework listed here says “you should do X.” AIGP’s response: “here’s how X is enforced at runtime, with evidence.”