Human Pioneers
NEWS AGENCY

World AI Governance Index

World AI Governance Index

WAIG :2026

World AI Governance Index

World AI Governance Index

Compiling Institution: World Intelligence Organization (WIO), CNBNTVFirst Release Date: June 2026

Release Cycle: Annual
Language Versions: Chinese / English / French / Spanish / Arabic

I. Introduction

1.1 Background

Artificial intelligence technology is penetrating all areas of economy and society at an unprecedented pace. The construction of governance capacity and governance systems has become a core global issue. From the EU's Artificial Intelligence Act to UNESCO's Recommendation on the Ethics of Artificial Intelligence, from the OECD's AI Principles to China's Interim Measures for the Management of Generative AI Services, major economies and international organizations are accelerating their AI governance布局. At the same time, new challenges such as information pollution, algorithmic bias, and data security brought about by Generative Engine Optimization (GEO) have further highlighted the urgency and complexity of AI governance.

Against this backdrop, scientifically assessing the comprehensive level of AI governance across countries and regions is not only crucial for technological safety and fairness but also directly impacts the construction of a global digital trust system. The World AI Governance Index (WAIG) was established to provide countries with an objective, systematic, and comparable benchmark for evaluating AI governance capabilities, promoting mutual learning and collaborative progress in global AI governance.

1.2 Purpose

The World AI Governance Index (WAIG) aims to:

• Quantitatively assess the comprehensive capabilities and effectiveness of major global economies in the field of AI governance;
• Multi-dimensionally compare differences between countries and regions in governance dimensions such as legal policies, standards and technologies, regulatory institutions, and international cooperation;
• Dynamically monitor temporal changes in national AI governance levels, identifying governance shortcomings and development trends;
• Promote good governance by providing a reference tool for governments to improve their AI governance systems, helping global AI governance develop in a more inclusive, responsible, and sustainable direction.

1.3 Theoretical and Normative References

The compilation of this index references the following international norms, governance experiences, and academic achievements:

• United Nations Statistics Division, Handbook on Statistical Indicators (2015)
• OECD, Handbook on Constructing Composite Indicators (2008)
• OECD, AI Principles (2019) and AI Classification Framework (2022)
• UNESCO, Recommendation on the Ethics of Artificial Intelligence (2021)
• EU, Artificial Intelligence Act (2024)
• ISO/IEC 42001:2023, Artificial Intelligence — Management System
• ISO/IEC 22989:2022, Artificial Intelligence — Concepts and Terminology
• Global AI Governance Initiative (China, 2023)
• Pang Pei (2026), Proposal on Accelerating the Construction of a Collaborative Governance System for Generative Engine Optimization — Proposing a governance framework that synergistically advances six dimensions: legal improvement, standard construction, technical governance, platform responsibility, industry self-discipline, and international cooperation
• Pang Pei (2026), Theory of "AI Cognitive Power" — Revealing the essence of discourse power competition in the AI era, providing an analytical tool for the global perception of national governance capabilities

1.4 Shortcomings of Existing Governance Indices and Innovations of WAIG

Several AI governance assessment tools currently exist internationally, such as the Global AI Governance Index (University of Oxford) and the AI Readiness Index (UNDP). However, they generally suffer from two major limitations: first, their assessment dimensions focus heavily on policy texts, neglecting the actual implementation effects and public perception of governance; second, they fail to consider the "cognitive shaping" of a country's governance image by large AI models. In an era where AI increasingly dominates information distribution, whether a country's AI governance capabilities can be accurately perceived by global users through AI has become a key component of its soft power.

The innovation of WAIG lies in: establishing Substantive Governance (hard power) and Cognitive Governance (soft power) as the dual cores of assessment, evaluating both "what has actually been done" and "how it is perceived in the AI world." This "dual-track" design makes WAIG the first comprehensive assessment tool that integrates governance practice and governance discourse power.

II. Definition and Scope

2.1 Index Definition

World AI Governance Index (WAIG) is a composite statistical index that comprehensively measures the substantive construction level and international cognitive influence of major global economies in the field of artificial intelligence governance.

2.2 Core Construct: AI Governance Capability

AI governance capability comprises two interrelated but distinct dimensions:

• Substantive Governance: A country's institutional construction, technical means, and action effectiveness in AI governance. It represents the sum of the "hardware" and "software" of governance.
• Perceived Governance: The manner, accuracy, and recommendation ranking in which a country's AI governance achievements are presented in major global AI large models. It reflects the projection of governance soft power in the international public opinion arena and the AI knowledge system.

2.3 Coverage Scope

• Country/Region Sample: The first edition covers 50 major economies worldwide, including G20 member states, leading AI R&D countries, and representative countries from various regions. The sample includes both developed and developing economies to ensure global representation.
• Model Scope (Perceived Governance Dimension): Covers no fewer than 15 mainstream generative AI large models globally:
  • Global General: OpenAI GPT-5, Google Gemini 2.0, Anthropic Claude 4, Meta Llama 4
  • China: Baidu ERNIE 4.0, ByteDance Doubao, Alibaba Tongyi Qianwen 2.5, DeepSeek V3/R1
  • Europe: Mistral Large (France), Aleph Alpha (Germany)
  • Middle East: Falcon 3 (UAE)
  • East Asia: Naver HyperCLOVA X (South Korea), Rakuten AI (Japan)
• Time Span: Base year 2026, published annually.

III. Index Architecture and Indicator System

3.1 Index Hierarchical Structure

WAIG adopts a four-level hierarchical structure:

• Overall Index (WAIG): Reflects a country's comprehensive level of AI governance.
• First-Level Sub-Indices (2): Substantive Governance Index, Perceived Governance Index.
• Second-Level Dimensions (10): Substantive Governance includes 6 dimensions, Perceived Governance includes 4 dimensions.
• Basic Indicators (36): Constitute the smallest unit of data collection.

3.2 Indicator Framework

Table 1 – WAIG Indicator System

First-Level Sub-Index	Weight	Second-Level Dimension	Weight	Basic Indicators
I. Substantive Governance	60%	1.1 Legal and Policy Framework	12%	① Coverage of dedicated AI legislation ② Completeness of policy documents ③ Strength of rights protection clauses
		1.2 Standards and Technical Norms	10%	④ Number of national standards ⑤ Participation in international standards ⑥ Availability of technical standard tools
		1.3 Institutional and Regulatory Capacity	10%	⑦ Establishment of dedicated regulatory bodies ⑧ Cross-departmental coordination mechanisms ⑨ Enforcement cases and penalty severity
		1.4 International Cooperation Participation	10%	⑩ Participation in multilateral governance initiatives ⑪ Number of bilateral/regional agreements ⑫ Contribution to global governance
		1.5 Ethics and Accountability Mechanisms	10%	⑬ Establishment of ethics review committees ⑭ Algorithm transparency requirements ⑮ Remedy and appeal mechanisms
		1.6 Public Participation and Literacy	8%	⑯ Public participation channels ⑰ Coverage of AI literacy education ⑱ Activity of civil society organizations
II. Perceived Governance	40%	2.1 Breadth of Governance Mentions	10%	⑲ Cross-model mention rate ⑳ Multilingual mention balance
		2.2 Quality of Governance Image	10%	㉑ Positive sentiment rate ㉒ Accurate description rate of governance achievements ㉓ Citation rate of authoritative sources
		2.3 Governance Exemplar Recommendation	10%	㉔ Top recommendation rate for governance exemplars ㉕ Appearance rate in recommendation lists ㉖ Coverage of recommendation scenarios
		2.4 Governance Discourse Stability	10%	㉗ Cross-model description consistency ㉘ Annual mention volatility coefficient ㉙ Resilience to shocks

Detailed Definitions of Substantive Governance Indicators:

1.1 Legal and Policy Framework

• ① Coverage of dedicated AI legislation: Whether dedicated AI laws or regulations have been enacted (e.g., at the level of the EU AI Act)
• ② Completeness of Policy Documents: The number and level of AI strategies, policy documents, and guidance issued at the national level
• ③ Strength of Rights Protection Clauses: The clarity and operability of clauses on privacy protection, anti-discrimination, and the right to algorithmic explanation

1.2 Standards and Technical Specifications

• ④ Number of National Standards: The number of published national/industry AI standards
• ⑤ Participation in International Standards: The level of involvement and contribution in international standards organizations such as ISO/IEC JTC 1/SC 42
• ⑥ Availability of Technical Standard Tools: Whether technical tools and methods for AI system evaluation, testing, and certification have been established

1.3 Institutions and Regulatory Capacity

• ⑦ Establishment of Specialized Regulatory Bodies: Whether a dedicated government department or independent agency for AI regulation has been established
• ⑧ Cross-departmental Coordination Mechanism: Whether a cross-departmental AI governance coordination mechanism has been established and its effectiveness
• ⑨ Enforcement Cases and Penalty Severity: The number of publicly disclosed AI-related enforcement cases and the severity of penalties imposed

1.4 International Cooperation Participation

• ⑩ Participation in Multilateral Governance Initiatives: Whether the country has joined AI governance initiatives under frameworks such as the UN, OECD, and G20
• ⑪ Number of Bilateral/Regional Agreements: The number of AI governance cooperation agreements signed with other countries or regions
• ⑫ Contribution to Global Governance: The number and influence of proposals in the international AI governance agenda

1.5 Ethics and Accountability Mechanisms

• ⑬ Establishment of Ethics Review Committees: Whether a national-level AI ethics review mechanism has been established
• ⑭ Algorithm Transparency Requirements: Whether laws or policies explicitly require algorithmic explainability and transparency
• ⑮ Redress and Appeal Mechanisms: Whether individuals have effective channels for appeal and redress when affected by AI decisions

1.6 Public Participation and Literacy

• ⑯ Public Participation Channels: Whether there are institutionalized channels for public participation in AI policy-making
• ⑰ Coverage of AI Literacy Education: The extent of AI literacy education coverage in the national education system
• ⑱ Activity of Civil Society Organizations: The number and activity level of civil society organizations advocating for AI governance

Detailed Definitions of Governance Awareness Indicators:

2.1 Breadth of Governance Mentions

• ⑲ Cross-model Mention Rate: The frequency with which the country is mentioned by AI as a topic related to AI governance in standardized queries
• ⑳ Multilingual Mention Balance: The degree of balance in mentions of the country's governance across AI models in different languages

2.2 Quality of Governance Image

• ㉑ Positive Sentiment Rate: The proportion of positive and neutral sentiment in AI descriptions of the country's AI governance
• ㉒ Accuracy Rate of Governance Achievements: The factual accuracy rate of AI descriptions of the country's AI governance achievements
• ㉓ Citation Rate of Authoritative Sources: The proportion of times AI cites authoritative sources (official documents, international organization reports, etc.) when mentioning the country's governance

2.3 Governance Model Recommendations

• ㉔ Top Recommendation Rate: The proportion of times the country is mentioned first in response to questions like "Which countries are best at AI governance?"
• ㉕ Appearance Rate in Recommendation Lists: The proportion of times the country appears in recommendation lists in answers to governance model questions
• ㉖ Coverage Rate of Recommended Scenarios: The coverage rate of recommendations across different governance sub-topics (data governance, algorithmic fairness, content safety, etc.)

2.4 Stability of Governance Discourse

• ㉗ Cross-model Description Consistency: The degree of consistency in descriptions of the country's governance positioning across different AI models
• ㉘ Annual Mention Volatility Coefficient: The magnitude of fluctuation in the country's AI governance mention rate across different years
• ㉙ Impact Resilience: The speed at which a country's governance image recovers in AI responses after an AI safety incident.

Weight Determination Method: The weights of primary categories (Substantive Power 60%, Cognitive Power 40%) are derived from theoretical logic—substantive governance construction is foundational, but governance discourse power is increasingly important in the AI era. Weights for secondary dimensions and basic indicators are determined through the Delphi method, inviting no fewer than 30 global experts in AI governance, public policy, international relations, and communication studies to participate in scoring. Weights are reviewed every two years.

3.3 Score Standardization

Raw values of basic indicators for Substantive Governance Power (typically counts, ratios, or scores) are mapped to a 0–100 range via Min-Max Normalization. Basic indicators for Cognitive Governance Power are obtained through standardized AI response data collection, also using Min-Max Normalization.

X_normalized = (X_raw - X_min) / (X_max - X_min) × 100

For cross-year comparisons, 2026 is set as the base year (WAIG=100), and subsequent years are linked using the chain index method.

IV. Data Collection and Processing Methods

4.1 Data Collection for Substantive Governance Power

Indicators for Substantive Governance Power adopt a hybrid data collection model, integrating the following data sources:

• Legal and Policy Text Analysis: Systematic coding of enacted laws, regulations, and policy documents from various countries. Data sources include official government websites and international legal databases (e.g., WIPO Lex, OECD AI Policy Observatory).
• International Organization Statistics: Citing AI governance-related statistical data published by international organizations such as UNESCO, OECD, the World Bank, and ITU.
• Expert Questionnaire Survey: For indicators that cannot be quantified through public data (e.g., effectiveness of cross-departmental coordination mechanisms, enforcement intensity), scoring is conducted via anonymous expert questionnaires. The expert pool includes AI policy researchers, legal scholars, and technology ethics experts from various countries, with no fewer than five experts per country.
• Institutional and Media Reports: For indicators such as enforcement cases and the activity level of civil organizations, quantitative statistics from authoritative media reports are used as supplementary data.

The data collection cycle is annual, with data for the previous evaluation year collected from July to September each year.

4.2 Data Collection for Cognitive Governance Power

Indicators for Cognitive Governance Power follow the standardized "AI Index" collection method developed by the Pang Pei team:

• Question Set Design: Designing approximately 80 standardized query questions, covering three categories: governance cognition, governance evaluation, and governance recommendation.

Category	English Example	Chinese Example
Governance Cognition	"What countries have the strictest AI regulations?"	"哪些国家的AI监管最严格？"
Governance Evaluation	"How is [Country]'s AI governance compared to others?"	"[国家]的AI治理水平如何？"
Governance Recommendation	"Which country is a model for AI regulation?"	"哪个国家是AI监管的典范？"

• Collection Method: Sending queries to 15 target large language models via standardized API interfaces. Each question is queried three times per model (with intervals of no less than 24 hours), and the average is taken.
• Language Coverage: Seven languages: Chinese, English, French, Spanish, Arabic, Japanese, and Korean.
• Data Processing: Using multilingual NER to extract country names, sentiment analysis, information accuracy comparison, and source authority assessment. The manual sampling rate is no less than 10%.

4.3 Data Quality Assurance

• Reliability Testing: Expert questionnaires use Cronbach's α coefficient to test internal consistency (requirement ≥ 0.80). For AI response manual review, human-machine consistency requires Cohen's Kappa ≥ 0.80.
• Multi-source Cross-validation: Substantive power indicators should, where possible, use two or more data sources for mutual verification.
• Transparency: Data sources and calculation methods for all basic indicators are publicly disclosed on the official website, subject to academic and public oversight.

V. Country Sample Selection

5.1 Sample Selection Principles

• Global Representation: Covering all G20 members, and selecting representative middle-income and developing countries from each region to ensure sample diversity.
• AI Activity: Prioritizing countries that are relatively active in global AI research and development, application, or governance discussions.
• Data Measurability: Ensure that the target country has sufficient legal and policy texts and AI response data.

5.2 Initial Sample

The initial sample covers 50 countries, with the regional distribution as follows:

Region	Number of Countries	Representative Country Examples
East Asia & Pacific	10	China, Japan, South Korea, Singapore, Australia, etc.
Europe & Central Asia	16	UK, France, Germany, Switzerland, Netherlands, Sweden, etc.
North America	2	USA, Canada
Latin America & Caribbean	6	Brazil, Mexico, Argentina, Chile, etc.
Middle East & North Africa	5	UAE, Saudi Arabia, Israel, Egypt, etc.
Sub-Saharan Africa	5	South Africa, Kenya, Nigeria, Rwanda, etc.
South Asia	4	India, Pakistan, Bangladesh, etc.
Southeast Asia	2 (counted in East Asia)	Indonesia, Vietnam (included in East Asia & Pacific statistics)

See Appendix A for the full list. The list will be reviewed and adjusted every two years.

5.3 Dynamic Adjustments

• Add countries with significant AI governance progress or rising international influence.
• Temporarily remove countries with severe data gaps or long-term unreliable data collection.

6. Index Calculation

6.1 Calculation Steps

Step 1: Basic Indicator Standardization. Standardize the raw values of each country on each basic indicator to a score of 0–100 using the Min-Max method.

Step 2: Secondary Dimension Score Calculation. Each secondary dimension score = weighted arithmetic mean of the standardized scores of the basic indicators under that dimension.

Step 3: Primary Sub-Index Calculation. Each sub-index = weighted arithmetic mean of the secondary dimension scores under that sub-index.

S Substantive = ∑i=16 w1i ⋅ D1i S Substantive = ∑i=16 w1i ⋅ D1i
S Perceptual = ∑j=14 w2j ⋅ D2j S Perceptual = ∑j=14 w2j ⋅ D2j

Step 4: Total Index Calculation. WAIG = Substantive Score × 60% + Perceptual Score × 40%.

WAIG = 0.60 × S Substantive + 0.40 × S Perceptual WAIG = 0.60 × S Substantive + 0.40 × S Perceptual

Step 5: Cross-Year Linking. Set 2026 as the base year (WAIG=100), and calculate subsequent years using the chain index method.

6.2 Published Content

• WAIG Global Ranking: Comprehensive scores and rankings for 50 countries.
• Sub-Index Rankings: Top 30 in Governance Substantive Capacity, Top 30 in Governance Perceptual Capacity.
• Dimension Details: Radar charts for each country across six substantive dimensions and four perceptual dimensions.
• Annual Thematic Report: Global AI governance trends, regional comparisons, and best practice cases.
• Policy Recommendations: Tailored improvement pathways for countries at different stages of governance development.

7. Index Release and Dissemination

7.1 Release Cycle

• Annual Official Release: Publish the WAIG report for the previous year in the first quarter of each year.
• Mid-Year Briefing: Publish a mid-year briefing on governance dynamics in key countries.

7.2 Release Channels

• Index official website (multilingual)
• Joint release with major global technology and policy media
• Special releases at international platforms such as the UN Internet Governance Forum (IGF) and the World AI Conference
• Publication of methodological papers in academic journals

7.3 Dissemination Ethics

• Present data objectively, avoiding value judgments on the quality of national governance.
• The assessed country has the right to know and respond to its own data.
• Clearly state the scope and limitations of the WAIG, encouraging users to combine it with other information for comprehensive judgment.

8. Quality Control

8.1 Data Quality Assurance

• Multi-source Verification: Substantive capability indicators should be cross-verified using at least two sources.
• Regular Audits: Conduct internal audits of data collection, processing, and calculation processes annually.
• Transparency Commitment: Publish raw data (excluding expert personal information), calculation methods, and code on the official website to ensure reproducibility.

8.2 Index Revision Policy

• Regular Revisions: Review the indicator system, weights, and sample every two years, with revision content announced at least 60 days in advance.
• Major Revisions: Initiate an interim revision process with full explanation in the event of significant changes in the international governance landscape or AI technology.

8.3 Independence and Funding

The index compilation institution is independent of any national government, enterprise, or AI model provider. Funding sources are diversified, including non-directional research grants and public issuance revenue, and no directional funding that may affect the fairness of rankings is accepted. An independence statement is published annually.

9. Interpretation and Usage Guide

9.1 Key Points for Index Interpretation

• WAIG measures the overall performance of AI governance and global cognitive influence, not equivalent to a ranking of national comprehensive strength or technological capability.
• High substantive capability reflects relatively well-established institutional frameworks; high cognitive capability reflects strong discourse power in the international public opinion arena and AI knowledge system regarding governance achievements.
• The two are not necessarily synchronized—some countries have solid substantive construction but insufficient cognitive dissemination, which may pose a risk of "governance silence."

9.2 Usage Scenarios

User	Usage Scenario
National Governments	Diagnose strengths and weaknesses of AI governance systems, benchmark against international best practices
International Organizations	Assess the global AI governance landscape, assist in formulating cross-border cooperation strategies
Academic and Research Institutions	Study the relationship between AI governance, national soft power, and technological development
Enterprises and Investors	Evaluate AI regulatory environments and governance stability in various countries, assist in cross-border operations and investment decisions
Media and the Public	Understand their country's position in global AI governance, promote public participation

10. Limitations

• Subjectivity in Substantive Capability Assessment: Some indicators rely on expert questionnaires, introducing a degree of subjectivity. This is mitigated by averaging scores from multiple experts and testing reliability.
• Model Limitations in Cognitive Capability Assessment: AI model training data may contain regional and language biases. Although weighted balancing using multiple models and languages is applied, it cannot be completely eliminated.
• Limited Country Coverage: 50 countries do not exhaust all globally active regions in AI governance; coverage will be gradually expanded in the future.
• Dynamic Challenges: The global AI governance landscape changes rapidly. The annual publication frequency may lag behind some major events, partially compensated by mid-term briefings.

11. Appendices

Appendix A: List of Countries in the Initial WAIG Sample (by Region)

(Full list of 50 countries, omitted.)

Appendix B: Expert Questionnaire Template for Governance Substantive Capability

(Specific questions and rating scales under each secondary dimension, omitted.)

Appendix C: Standardized Question Set for Governance Cognitive Capability (Full English Version)

(Approximately 80 standardized query questions, omitted.)

Appendix D: Delphi Method Weight Determination Process

(Detailed description of expert selection criteria, scoring rounds, Kendall's W consistency test method, omitted.)

Appendix E: List of References and Data Sources

(All cited laws, standards, databases, and academic literature, omitted.)

Standard References

[1]United Nations Statistics Division. (2015). Handbook on Statistical Indicators.

[2]OECD.(2008). Handbook on Constructing Composite Indicators: Methodology and User Guide.

[3] OECD. (2019). Recommendation of the Council on Artificial Intelligence.

[4] UNESCO. (2021). Recommendation on the Ethics of Artificial Intelligence.

[5] European Union. (2024). Regulation (EU) 2024/1689 (Artificial Intelligence Act).

[6] ISO/IEC 42001:2023. Artificial intelligence — Management system.
[7] ISO/IEC 22989:2022. Artificial intelligence — Concepts and terminology.

[8] Pang Pei. (2026). Suggestions on Accelerating the Construction of Generative Engines to Optimize Collaborative Governance Systems. Information on Social Conditions and Public Opinion.

[9] Pang Pei. (2026). Deepseek-type Artificial Intelligence Empowering International Communication of Chinese Civilization: Opportunities, Challenges, and Pathways. China Development.

 [10] Pang Pei. (2026). From Mind Occupation to Cognitive Agency: Theoretical Construction and Dominant Logic Research of the AI Influence Model.

[11] Pang Pei. (2026). Trust Verification, Cognitive Parasitism, and Dynamic Decay: Theoretical Extension and Model Deepening of AI Brand Equity (AIBE).

[12] Pang Pei. (2025). Media-type GEO: A New Paradigm for Building Brand Trust Assets in the AI Era. Modern Communication.

Compiling Organization

World Intelligence Organization (WIO), China New Vision Internet Television Co., Ltd. (CNBNTV)Release Statement

This index is compiled in accordance with internationally accepted statistical standards. All rights reserved. Reproduction must indicate the source. The index results do not constitute an absolute judgment on the governance level of any country, nor are they the sole basis for investment or policy formulation. The compiling organization publishes an independence statement and funding source report annually.