International Telecommunication Union Annual AI Governance Report: Leading the Future of Artificial Intelligence

Key Chapter Title List

1. White Paper: Themes and Trends in AI Governance
2. Theme One: The Year of AI Agents
3. Theme Two: AI and Socioeconomic Impact
4. Theme Three: International AI Governance Coordination
5. Theme Four: AI Standards and Best Practices
6. Theme Five: AI Infrastructure and Computing Power
7. Theme Six: AI Safety and Risk Management
8. 2025 AI Governance Dialogue Highlights and Insights
9. Global Context: Opportunities and Risks
10. Ten Pillars of AI Governance
11. Regional Perspectives and Case Studies
12. Vision Towards 2026

Document Introduction

This report is the flagship 2025 annual AI governance publication released by the International Telecommunication Union (ITU), produced in collaboration with institutions such as the Oxford Martin School. The report integrates a comprehensive trends white paper with the outcomes of a high-level AI governance dialogue that brought together stakeholders from over 169 countries and more than ten thousand participants. It aims to provide evidence-based in-depth analysis and practical guidance for the complex landscape of global AI governance.

The white paper in the first part of the report systematically reviews key developments from 2024 to early 2025, identifying seven core themes. The report notes that the AI field is transitioning from chat tools into the "Year of AI Agents," where systems capable of autonomously executing multi-step tasks present novel governance challenges, including traceability, security vulnerabilities, and liability definition. At the socioeconomic level, AI is reshaping labor markets, with its impact presenting a complex picture of simultaneous automation and job restructuring, while also fostering new value chains in fields such as education, science, transportation, and agriculture. However, global AI development is highly uneven, with critical infrastructure like computing power and data centers heavily concentrated in the United States, China, and the European Union. Over 150 countries lack advanced computing hubs, exacerbating the "AI divide."

At the international governance level, the report analyzes how the series of global AI summits from Bletchley, Seoul to Paris have evolved from ad-hoc meetings into structured cooperation platforms. It also explores the progress made by Sino-US "Track II" diplomacy and regional partnerships (such as ASEAN, the African Union, and the G7) in building AI safety norms and cooperation agreements. Simultaneously, the report provides a detailed examination of the global AI standards landscape led by bodies like ITU, ISO, and IEC, as well as the development and existing gaps in ethical frameworks, safety testing (e.g., red teaming), and certification programs.

The second part of the report focuses on the essence of the 2025 ITU AI Governance Dialogue. The dialogue revealed core tensions in global AI governance: the "pace problem" of rapid technological evolution versus lagging policy-making; the risk of power imbalance due to the high concentration of power (computing power, data, talent) in a few countries and companies; and the inequality and emerging digital divide exacerbated by disparities in connectivity, access to computing power, and language representation. Case studies from countries like Estonia, Switzerland, Singapore, Saudi Arabia, Zimbabwe, and Costa Rica demonstrate how nations of different sizes and endowments are adopting differentiated strategies (e.g., Estonia's comprehensive digitalization and AI application, Switzerland's open multilingual models and the International Computing Alliance and Innovation Network - ICAIN) to enhance AI sovereignty and inclusivity.

Based on extensive discussions, the report ultimately distills the "Ten Pillars of AI Governance" proposed by the dialogue's co-chairs, serving as input for the upcoming United Nations "Global Dialogue on AI Governance" set to launch in 2026. These ten pillars cover key areas of action ranging from principles to practice, multi-stakeholder participation, transparency, inclusivity, capacity building, environmental sustainability, industry focus, standards and safety tools, computing power and model governance, to policy interoperability and agile governance. They provide a systematic blueprint for building a global governance framework capable of both managing risks and unleashing AI's potential.