⚡ Quick Summary
Published by the Center for Long-term Artificial Intelligence (CLAI), Beijing Institute of AI Safety and Governance (Beijing-AISI), Beijing Key Laboratory of Safe AI and Superalignment, and the International Research Center for AI Ethics and Governance at the Institute of Automation, Chinese Academy of Sciences, this pre-release report presents a cross-country AI governance evaluation. It assesses 40 countries across income groups, regions and technological development stages, expanding the preceding edition’s coverage from 14 countries and its indicator set from 39 to 43.
The index is built on the principle that “the level of governance should match the level of development.” Its framework has four pillars—AI Development Level, AI Governance Environment, AI Governance Instruments, and AI Governance Effectiveness—divided into 17 dimensions and 43 indicators. It combines policy documents, governance practices, research outputs and risk-exposure data into comparable country scores. The report groups countries into three overall score tiers and four profiles: All-round Leaders, Governance Overachievers, Governance Shortfallers and Foundation Seekers. It also supplies an indicator catalogue, source notes, scoring rules, normalisation and missing-data procedures intended to support repeated comparison over time.
🧩 What’s Covered
The report moves from its conceptual model and aggregate findings to pillar-level observations, then provides the underlying indicator and scoring methodology.
- Framework and scope: Defines AGILE as a multi-layered index intended to align governance maturity with technological development. It introduces the four pillars, 17 dimensions and 43 quantifiable indicators, and explains that the 2025 edition adds country coverage, data sources, generative-AI data and a revised missing-data imputation approach.
- Overall results: Describes three score tiers among the 40 assessed countries, comparatively wide variation in AI Development Level and AI Governance Instruments, and a generally positive relationship between total score and GDP per capita. It distinguishes four country types based on performance across the pillars.
- AI Development Level: Covers AI research and development activity, infrastructure and industry vitality. Measures include AI publications, active researchers, granted patents, large-scale AI systems, data centres, supercomputing, ICT development, private investment and newly funded AI companies. It also reports trends in generative-AI patents.
- AI Governance Environment: Examines AI risk exposure and overall governance readiness. The report draws on recorded AI incidents and assesses general governance, digital development and sustainable development conditions. It highlights reported growth in incidents and the prominence of robustness and digital security, human-rights, and privacy and data-governance risks.
- AI Governance Instruments: Reviews national AI strategies, governance bodies, principles and norms, impact assessments, standards and certification, legislation, and international engagement. It explains the treatment of comprehensive laws, sector-specific rules and AI-related data or information protection provisions.
- AI Governance Effectiveness: Assesses public understanding, social acceptance, development inclusivity, data and algorithm openness, and AI governance research activity. The indicators cover AI literacy and attitudes, gender and digital inclusion measures, influential open models and datasets, developer contributions, and publications on governance, safety and security, and AI for Sustainable Development Goals.
- Appendix and methodology: Lists indicator-level sources and links dimensions to UNESCO’s Recommendation on the Ethics of Artificial Intelligence and Readiness Assessment Methodology. It explains literature classification, binary strategy scoring, legislative scoring, averaging, standardisation, percentile-fit normalisation and hierarchical imputation.
💡 Why it matters?
The report gives government and policy teams a structured way to compare AI governance conditions with AI development, rather than treating national AI capacity as a single measure. Its separate treatment of instruments, risk exposure, public understanding, inclusion and openness can help users identify where a country’s policy infrastructure or enabling conditions differ from its technical development.
It is also useful for reviewers tracing how national strategies, legislation, impact assessment mechanisms, standards, certification and international participation are operationalised as index indicators. The mapping to UNESCO recommendation articles and the indicator on ISO AI standardisation participation make the framework legible alongside those international governance reference points.
❓ What’s Missing
The index relies on heterogeneous source periods and methods, including desk research, external indices, platform data and literature analysis. The report describes imputation where current or historical indicator data are absent, so some results may include estimated values rather than directly observed data. Its literature method attributes nationality from institutional addresses and, where unavailable, collaboration networks; it infers gender from names using country-specific conventions and acknowledges ambiguity in gender-neutral or culturally variable names. Governance-publication classification also uses venue and title-keyword filtering. The supplied extraction does not contain readable country-profile material on pages 45–59 or the numerical content of several referenced tables and figures, limiting assessment of country-level evidence in this review.
👥 Best For
This report is best suited to public-sector AI governance teams, researchers and international-policy analysts comparing national AI capacity across development, risk, governance tools and observed effectiveness. It also serves teams designing country-level benchmarks or reviewing AI strategies, legislation, impact assessment mechanisms and international AI governance participation.
📄 Source Details
AI Governance InternationaL Evaluation Index (AGILE Index) 2025 is an English 2025 pre-release report, version v1.0.0-pre, made available for preview on 8 July 2025. The citation credits Yi Zeng, Enmeng Lu, Xiaoyang Guo, Cunqing Huangfu, Jiawei Xie, Yu Chen, Zhengqi Wang, Dongqi Liang, Gongce Cao, Jin Wang, Zizhe Ruan, Xin Guan and Ammar Younas, alongside the four listed institutes. The report website is https://agile-index.ai/. The supplied extraction covers pages 1–80 of an 81-page PDF.
