If your analytical framework treats G20 AI policy activity as broadly similar across member countries, the distribution of enacted AI legislation from 2016 to 2025 is uneven by an order of magnitude, and the methodological treatment of legislative volume requires careful interpretation. The US passed 25 AI-related federal bills since 2016. South Korea 17. Japan and France 10 each. Italy 9. Germany and the UK 6 and 5 respectively. China and India 5 and 3. Russia and Saudi Arabia passed very few or no AI-specific federal legislation. The G20 legislative output picture has substantial inter-country variance that needs methodological interpretation before strategic conclusions can be drawn.

The methodological framing matters. Three things to note about the count.

First: the dataset covers enacted legislation only, not proposed or pending bills. Many AI-relevant policy frameworks operate through regulation, executive action, or implementation of pre-existing legislation rather than through new dedicated AI bills. The enacted-legislation count can therefore understate the actual volume of AI-related policymaking in countries that operate through other policy instruments.

Second: large omnibus bills containing multiple AI provisions are counted as a single piece of legislation. A country that bundles AI provisions into broader legislation will have a lower bill count than a country that issues stand-alone AI bills for each issue, even if the substantive policy coverage is comparable.

Third: volume is not a measure of significance. A single major law (the EU AI Act, for example, though it is EU-level rather than national) can carry more impact and enforcement weight than dozens of narrower national-level bills. The 25 US bills include the federal Take It Down Act (May 2025), itself a significant national framework, alongside many narrower provisions. The 17 South Korean bills include the Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust, a foundational comprehensive framework. The legislative weight is not the same as the bill count.

With those methodological caveats noted, the cross-country pattern is informative.

The US leads in cumulative count at 25 federal bills since 2016. This count specifically covers federal legislation; the US state-level count is separately tracked at 482 bills cumulatively (substantially more than any other country's combined federal and subnational count). The US federal-plus-state framework is the most active legislative AI environment among G20 countries.

South Korea's 17 bills reflect a concentrated AI legislative programme that includes the Framework Act, sector-specific bills, and operational provisions. South Korea's legislative process can pass AI bills relatively quickly when policy alignment exists; the 17-bill count partly reflects this throughput.

Japan, France, and Italy in the 9-10 range represent moderate legislative activity with substantive frameworks. Japan's Act on the Promotion of Research and Development and Utilization of Artificial Intelligence–Related Technology, Italy's Law No. 132/2025 (the first EU member state national AI law), and France's various AI provisions all carry significant policy weight.

Germany at 6 and the UK at 5 reflect a different policy approach. Both countries operate substantial AI policy frameworks through regulation, sectoral law, and implementation of EU-level requirements (for Germany) or post-Brexit policy machinery (for the UK). The lower bill count understates the actual policy activity.

China at 5 and India at 3 reflect both genuine lower bill output and the methodology issue described above: both countries operate substantial AI policy through regulatory and executive instruments that may not appear as "AI-related bills passed into law." China's mandatory AI-generated content labelling rules (finalised March 2025) operate as a regulatory framework rather than primary legislation. India's various AI initiatives operate through executive and regulatory mechanisms. The bill count understates actual policy activity for both countries.

Russia and Saudi Arabia at near-zero legislative output reflect both lower priority for formal AI legislation and a heavier reliance on executive and administrative instruments. Russia's AI strategy operates through executive policy direction; Saudi Arabia's AI activity centres on the country's Vision 2030 framework and Saudi Data and Artificial Intelligence Authority (SDAIA) rather than legislative bills.

The methodological observation: the cross-country comparison of G20 AI legislative activity is informative only when paired with the methodology of how each country pursues AI policy. The count alone can mislead. The strategic interpretation requires understanding which countries operate primarily through legislation, which operate through regulation, and which operate through executive policy direction.

Three structural observations follow from the data with the methodological caveats applied.

The first observation: legislative activity is expanding but unevenly across G20. The 2016 baseline was zero AI-related bills across the G20. By 2025, every G20 country except a few has at least some AI legislative activity. The expansion is real even if the distribution is uneven. The trajectory points to more, not less, AI-specific legislation across G20 countries in 2026-2030.

The second observation: countries that pursue AI policy through legislation (US, South Korea, Italy, Japan, France) produce more visible AI policy signal than countries that pursue it through regulation or executive action (Germany, UK, China, India). The visibility difference can affect international perception of national AI policy maturity, but the underlying policy activity may be comparable. Organisations evaluating national AI policy maturity should engage with the actual policy framework rather than the bill count alone.

The third observation: the methodology of counting "AI-related bills" depends on definitional choices that vary across studies. The 2026 counting methodology has changed from earlier years (counts may differ from figures reported in prior years). Bills are counted based on AI relevance, which itself requires interpretation. The Digital Policy Alert methodology used for the 2026 data is the most systematic to date but still requires judgement calls. Cross-year comparisons within a single methodology are more reliable than comparisons across methodologies.

For organisations operating across G20 jurisdictions, the methodological recommendation is to engage with each country's actual AI policy framework rather than rely on the bill count as a single metric. The bill count is one data point among many: regulation count, executive policy direction, sectoral law application, and international agreement participation all add to the picture. A multi-metric national AI policy assessment produces more accurate strategic positioning than the bill count alone.

The methodological framing: the G20 AI legislation count is informative as a directional signal, confirming that AI is a globally active policy area with growing activity across most major economies, but is methodologically constrained as a comparative metric. The 25-1 ratio between the US and Saudi Arabia in bill count does not mean the US has 25x more AI policy activity than Saudi Arabia. It means the two countries have chosen substantially different policy instruments and methodologies. Strategic plans that read the bill count as the primary signal of national AI policy maturity will be reading a partial signal.

Sources

  • Primary: Stanford AI Index 2026, Chapter 8 (Policy and Governance) 8.4 — hai.stanford.edu/ai-index/2026
  • Legislative tracking: Digital Policy Alert, 2026 — G20 AI legislation tracking, 2026 methodology
  • Cross-validation: Stanford AI Index 2026 manual review of inclusions and exclusions
  • Country-specific references: US Take It Down Act (May 2025); South Korea Framework Act on AI; Italy Law No. 132/2025; Japan Act on the Promotion of Research and Development and Utilization of AI-Related Technology; China AI-generated content labelling rules (March 2025); Saudi Arabia Vision 2030 / SDAIA framework