In 2024, more than half of newly adopted national AI strategies came from emerging economies — a marked shift in AI governance from earlier years, when high-income countries dominated. As of 2025, additional countries across sub-Saharan Africa, Central Asia, and the Middle East have strategies in active development.

New frameworks at the edges

• Sub-Saharan Africa surfaced new strategies in Ethiopia, Ghana, and Nigeria, while South Africa has one already in development.
• South and Central Asia saw frameworks emerge in Sri Lanka and Nepal; Latin America and the Caribbean added Costa Rica and Jamaica, with Mexico's in development.
• High-income economies kept contributing, but more slowly and with a focus on consolidation — European countries such as Malta released updated strategies to align with EU AI Act requirements.

As more countries adopt strategies, a consensus is rising that AI can be a lever to bolster state capacity. The next challenge, the report notes, is implementation and regulatory capacity — particularly in Africa, where many countries still lack formal strategies and risk falling behind in AI governance and readiness.

Compute: a sevenfold gap

Between 2018 and 2025, Europe and Central Asia expanded state-backed AI supercomputing clusters from 3 to 44, largely driven by the European High Performance Computing Joint Undertaking (EuroHPC JU). North America grew nearly sevenfold to 41 clusters, and East Asia (excluding China) about fourfold to 27. By contrast, South Asia, the Middle East and North Africa, and Latin America and the Caribbean each only doubled or tripled — reaching 2, 3, and 8 clusters respectively by 2025.

Data: three regional clusters

• Through 2024, East Asia and the Pacific had adopted 77 data-localization measures, followed closely by sub-Saharan Africa (71) and Europe and Central Asia (66).
• Moderate-localization regions include the Middle East and North Africa (44), Latin America and the Caribbean (36), and South Asia (24).
• North America remains a striking outlier at just 3 measures, reflecting a long-standing 'flow-first' policy orientation that favors free cross-border data flows.

Model production remains concentrated, with the United States and China accounting for a disproportionate share of global activity, even as open-source development diffuses more broadly. Private firms — through OpenAI's Stargate and Nvidia's 'AI Factory' partnerships — are increasingly central to building what many governments designate as national AI infrastructure.

Among G20 countries, the United States passed the most AI-related bills between 2016 and 2025 (25 in total), followed by South Korea (17). At the state level the totals are far larger: California enacted 62 AI-related bills over the period, more than double any other state, followed by Maryland (28), Virginia (25), and Utah (24). Two states — Missouri and Rhode Island — have enacted none.

Federal reversal

Early in 2025 the federal direction changed: the administration rescinded Biden's Executive Order 14110 and issued 'Removing Barriers to American Leadership in Artificial Intelligence,' reorienting policy toward fewer constraints. Federal AI-related regulations still rose to 58 in 2025 (from one in 2016), with the Executive Office of the President alone issuing 28. In December 2025, a further executive order directed the Department of Justice to set up an AI Litigation Task Force to challenge state AI laws in court.

States fill the vacuum

In the absence of a federal framework, states moved ahead on their own — with targeted protections against discrimination, misinformation, and abuse, plus chatbot disclosure rules, watermarking and provenance requirements, and even a 'right to compute.' The US Senate also struck a proposed 10-year federal moratorium on state AI regulation from a major spending bill, allowing states to keep legislating.