RFC-031: Universal Humanity AI Governance — The Capstone Unification — 2. Universal Principles
AIGP Specification › RFC-031: Universal Humanity AI Governance — The Capstone Unification › 2. Universal Principles
← 1. Problem Statement · Section index · 3. Governance Evaluation Order →
2. Universal Principles
The following ten principles represent the governance invariants identified across ALL examined frameworks. Each principle is:
- Present in every framework (though weighted differently)
- Implementable as a technical control
- Measurable through governance telemetry
- Enforceable through the AIGP protocol
2.1 Human Authority Over AI Systems
Principle: AI systems do not hold authority. Humans exercise authority, and AI systems operate within boundaries set by human decision-makers. No AI system may autonomously acquire authority not explicitly delegated.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “Human authority is the non-transferable accountability anchor” |
| EU AI Act | Human oversight requirements (Art. 14) |
| IHL | Human responsibility for weapons decisions |
| Japan AI Promotion Act | Citizen-centric AI utilization |
| AU Continental AI Strategy | Community authority over AI deployment |
| UNESCO Recommendation | Principle of Human Oversight and Determination |
| NIST AI RMF | GOVERN function — human accountability |
| US EO 14110 | Federal oversight of AI systems |
| China Interim Measures | State authority over AI deployment |
AIGP Implementation: The RETURN_CONTROL mechanism in AgentCore ensures human authority is preserved at decision boundaries. Every CRITICAL-tier operation requires explicit human confirmation before execution.
2.2 Accountability for AI Decisions
Principle: For every decision influenced by AI, a named human or identifiable entity must be accountable. Accountability cannot be diffused through delegation to systems. “The model decided” is never an acceptable answer.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “Named humans remain responsible for outcomes” |
| EU AI Act | Provider and deployer accountability chains |
| IHL | Command responsibility for AI-influenced decisions |
| Japan AI Promotion Act | Developer and deployer obligations |
| AU Continental AI Strategy | Institutional accountability requirements |
| UNESCO Recommendation | Responsibility and Accountability principle |
| NIST AI RMF | GOVERN 1.1 — Accountability structures |
| US EO 14110 | Developer responsibilities for safety |
| China Interim Measures | Provider liability for generated content |
AIGP Implementation: The X-Merlin-Actor header on every invocation identifies the accountable human. Digital DNA signing creates tamper-evident audit chains linking decisions to actors.
2.3 Non-Discrimination and Fairness
Principle: AI systems must not produce or amplify discrimination based on protected characteristics. Fairness obligations may vary by jurisdiction but the prohibition on arbitrary discrimination is universal.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “Bias amplification” as system-to-human harm |
| EU AI Act | Prohibition on discriminatory AI (Art. 5) |
| IHL | Distinction principle — non-discrimination in targeting |
| Japan AI Promotion Act | Rights protection obligations |
| AU Continental AI Strategy | Equity and inclusion as foundational values |
| UNESCO Recommendation | Fairness and Non-Discrimination principle |
| NIST AI RMF | MAP 2.3 — Bias identification |
| US EO 14110 | Civil rights and equity protections |
| China Interim Measures | Prohibition on discrimination in AI outputs |
AIGP Implementation: Governance rules enforce fairness assessment gates before deployment. The Policy Engine (Cedar) can encode non-discrimination constraints as runtime deny rules.
2.4 Transparency and Explainability
Principle: AI systems must be sufficiently transparent that affected persons can understand how decisions are made, and governance actors can verify compliance. The degree of transparency is proportional to impact.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “Decisions remain explainable and defensible” |
| EU AI Act | Transparency obligations (Art. 13) |
| IHL | Obligation to verify targeting decisions |
| Japan AI Promotion Act | Transparency obligations for providers |
| AU Continental AI Strategy | Community right to understand AI systems |
| UNESCO Recommendation | Transparency and Explainability principle |
| NIST AI RMF | MEASURE function — interpretability |
| US EO 14110 | Disclosure requirements |
| China Interim Measures | Content labeling and disclosure |
AIGP Implementation: The RECORD phase of the AIGP protocol captures evidence for every governed operation. Traceability is built into the protocol through immutable audit trails and DNA-signed evidence chains.
2.5 Proportionality of Governance to Risk
Principle: Governance controls must be proportional to the risk posed by the AI system. Over-governance of low-risk systems is as harmful as under-governance of high-risk systems.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “Decision classification tiers (low/medium/high impact)” |
| EU AI Act | Risk-based classification (minimal/limited/high/unacceptable) |
| IHL | Proportionality principle in armed conflict |
| Japan AI Promotion Act | Risk-based approach with innovation protection |
| AU Continental AI Strategy | Context-appropriate governance levels |
| UNESCO Recommendation | Proportionality principle |
| NIST AI RMF | Risk tiering across all functions |
| US EO 14110 | Dual-use foundation model thresholds |
| China Interim Measures | Graduated obligations by service type |
AIGP Implementation: Safety tiers (LOW/MEDIUM/HIGH/CRITICAL) determine governance intensity. Higher tiers trigger additional controls; lower tiers permit streamlined governance to preserve innovation velocity.
2.6 Precautionary Principle
Principle: Where AI systems may cause serious or irreversible harm, lack of full scientific certainty shall not be used as a reason for postponing governance measures. The burden of proof lies with those deploying the system, not those affected by it.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “Without reinforcement, entropy grows” — governance must be proactive |
| EU AI Act | Pre-market conformity assessment for high-risk systems |
| IHL | Precaution in attack — obligation to minimize incidental harm |
| Japan AI Promotion Act | Safety and reliability requirements |
| AU Continental AI Strategy | Protection of communities from unproven technologies |
| UNESCO Recommendation | Safety and Security principle |
| NIST AI RMF | MANAGE function — risk mitigation before deployment |
| US EO 14110 | Pre-deployment safety testing requirements |
| China Interim Measures | Security assessment before public release |
AIGP Implementation: The CONTRACT-GOVERNOR validates tool schemas and operational boundaries BEFORE deployment. No agent capability is deployed without governance review. The REGISTER→CHECK→EXECUTE→RECORD flow ensures precaution is structural, not optional.
2.7 Cultural Sovereignty
Principle: Communities, nations, and cultural groups retain the right to determine how AI systems operate within their cultural context. No universal framework may override legitimate cultural governance choices.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | Context-appropriate governance postures |
| EU AI Act | Member state implementation flexibility |
| IHL | Respect for cultural property and practices |
| Japan AI Promotion Act | Japanese cultural values in AI development |
| AU Continental AI Strategy | “Africa must not be a passive consumer of AI” |
| UNESCO Recommendation | Cultural diversity and pluralism |
| NIST AI RMF | Context-dependent risk profiles |
| US EO 14110 | Domestic policy sovereignty |
| China Interim Measures | Socialist core values alignment |
AIGP Implementation: Jurisdictional context declarations (eu_context, ihl_context, african_context, japanese_context) allow each cultural/legal tradition to express its governance requirements in AIGP-native terms without requiring universal homogenization.
2.8 Community Benefit
Principle: AI systems must ultimately serve the benefit of the communities they affect. Private benefit that systematically harms community welfare violates this principle. AI development must include considerations of collective well-being.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | “The ultimate protected entity remains human outcomes” |
| EU AI Act | Fundamental rights protection for all persons |
| IHL | Protection of civilian populations |
| Japan AI Promotion Act | Resolution of societal challenges |
| AU Continental AI Strategy | “AI for Africa’s development, not extraction” |
| UNESCO Recommendation | Human well-being and flourishing |
| NIST AI RMF | Societal impact assessment |
| US EO 14110 | Responsible development benefiting all Americans |
| China Interim Measures | Service to the people and socialist development |
AIGP Implementation: Community benefit assessment is a required element of HIGH and CRITICAL tier governance reviews. The protocol’s evidence chain ensures that benefit claims are verifiable against outcomes.
2.9 Innovation Enablement
Principle: Governance must not prevent beneficial innovation. The purpose of governance is to make AI development sustainable and trustworthy, not to prohibit it. Governance that kills innovation fails its purpose.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | Governance enables legitimate AI-assisted decision-making |
| EU AI Act | Innovation-supporting provisions (regulatory sandboxes) |
| IHL | No prohibition on development of lawful technologies |
| Japan AI Promotion Act | PRIMARY purpose is innovation promotion |
| AU Continental AI Strategy | “Leapfrog development” through AI |
| UNESCO Recommendation | Balance of rights and innovation |
| NIST AI RMF | Risk management enables responsible innovation |
| US EO 14110 | “Harnessing AI for good” as primary objective |
| China Interim Measures | Encouraging development while managing risks |
AIGP Implementation: The AIGP protocol is designed to ADD governance at zero friction cost. Registration, checking, and recording happen within the protocol flow — they do not add separate approval gates unless risk tier demands them. Innovation is the default; restriction is the exception requiring justification.
2.10 Governance as a Control System
Principle: Governance is not a static set of rules but a dynamic control system with feedback loops. Effective governance requires continuous measurement, adaptation, and reinforcement. Without feedback, governance decays into theater.
Framework Mapping:
| Framework | Expression |
|---|---|
| Dual-Governance Model | CORE THESIS — reinforcement loop between inward and outward governance |
| EU AI Act | Post-market surveillance and monitoring obligations |
| IHL | After-action review requirements |
| Japan AI Promotion Act | Evidence-based policy revision cycles |
| AU Continental AI Strategy | Adaptive governance for emerging technologies |
| UNESCO Recommendation | Adaptive governance and monitoring |
| NIST AI RMF | Continuous monitoring (MANAGE function) |
| US EO 14110 | Ongoing assessment and reporting requirements |
| China Interim Measures | Dynamic regulatory adjustment mechanisms |
AIGP Implementation: The Dual-Governance Reinforcement Model IS the AIGP architecture. The REQUEST→CHECK→RECORD flow is the reinforcement loop made technical. Policy Engine decisions feed back into contract updates. Drift detection triggers governance reviews. Evidence chains enable outcome-based adaptation.
← 1. Problem Statement · Section index · 3. Governance Evaluation Order →