⚡ Quick Summary
Published as part of AI Index Report 2026, this report chapter surveys policy and governance developments around the world. The supplied chapter does not name a publisher. It combines a chronology of major policy news in 2025 with comparative evidence on national AI strategies, AI sovereignty, legislation, US policy activity, and public investment. Its analysis draws on national-strategy databases, legislative records, congressional-witness data, Epoch AI, Zeki, Brookings, and procurement data from the United States and Europe.
The chapter frames AI sovereignty as a state’s capacity to make independent decisions over AI development, deployment, and governance. It examines five dimensions: infrastructure, data, models, applications, and talent. Its indicators show continuing concentration in advanced compute and model releases alongside broader policy adoption and selective national investment. The chapter also tracks enacted AI legislation across G20 countries, US state and federal activity, and public spending. It reports, among other findings, 150 AI-related bills passed by US states in 2025, 58 US AI-related regulations in 2025, and approximately $20.5 billion in US AI-related public investment from 2013 to 2024.
🧩 What’s Covered
The chapter proceeds from a global policy timeline to comparative measures of state capacity, legislation, and spending.
- Major global policy news: A dated timeline records events during 2025, including US executive orders, the first EU AI Act measures, Chinese rules on labelling AI-generated content, the UN’s scientific panel and global dialogue, national AI laws, and supply-chain cooperation.
- National AI strategies: Using Oxford Insights data, this section tracks countries with adopted or developing formal strategies by geographic area. It stresses that the dataset records published policy intent rather than whether strategies have been implemented effectively.
- AI sovereignty: The chapter defines sovereignty across infrastructure, data, models, applications, and talent. It compares public and public-private AI supercomputers, regional data-localization measures, publicly reported model releases, investment across application areas, and cross-border flows of top AI authors and inventors.
- Infrastructure, data, and model indicators: Figures show 85 public or public-private AI supercomputers in China in 2025, 44 in Europe and Central Asia, and 41 in North America. Data-localization measures through 2024 range from 77 in East Asia and the Pacific to 3 in North America, while model-release counts remain led by the United States and China.
- Global and US policymaking: The chapter counts enacted AI-related laws in G20 countries from 2016 to 2025, with short profiles of laws in the United States, Italy, Japan, and South Korea. It then examines US state legislation, congressional hearing witnesses, and regulatory activity by federal agency.
- Public investment: Procurement, grants, and Other Transaction Agreements are used to compare US and European public AI spending. The chapter distinguishes US obligations from European awarded amounts, identifies leading US funding agencies, and presents country and sector patterns in European contracts.
💡 Why it matters?
For public-sector leaders, policy teams, and organisations exposed to changing AI requirements, the chapter provides a common evidence base for locating governance activity alongside investments in compute, data, models, and skills. Its five-part sovereignty framing connects regulatory choices with practical dependencies on infrastructure, cross-border data, procurement, and talent.
The comparative measures also help readers interpret legislative totals cautiously. The chapter notes that counts do not measure a law’s significance, that omnibus legislation may be counted once, and that state tracking based on the phrase “artificial intelligence” captures only part of the wider policy picture.
❓ What’s Missing
The chapter is an overview rather than a guide to compliance with any individual regime. Its national-strategy dataset captures what governments have published, not implementation or effectiveness. Legislative figures are limited to enacted laws and, for the global charts, G20 countries; they can understate activity and do not indicate enforcement weight. The sovereign-compute measure is a proxy for high-end training infrastructure rather than all national compute resources. Model-release figures are conservative where reporting is less systematic, and smaller language-specific models in sub-Saharan Africa are not represented. European contract figures show awarded amounts, often maximum ceilings, rather than clear timelines or totals of actual obligations.
👥 Best For
Government policy teams comparing national AI capacity and regulatory activity; public-procurement and investment analysts; and governance leads tracking how data localization, compute infrastructure, model development, and US federal or state policy shape AI deployment. It is also useful for researchers seeking stated methodological cautions alongside the reported figures.
📄 Source Details
The supplied complete 37-page English-language PDF is the chapter Policy and Governance from AI Index Report 2026, spanning printed pages 323–359. It identifies the report year as 2026. No individual authors, publishing organisation, edition or version statement, or URL for this PDF are printed in the supplied chapter.
