How countries write their AI futures – mapping the many models of AI governance

“In penning their national AI strategies governments are not only deciding how to regulate AI. They are also defining what AI should deliver, from economic growth to public-sector transformation. A text-based analysis using x.Machina, the open-access governance mapping platform co-founded by Elise Antoine and Adam Chalmers , shows that these priorities combine in more varied ways than the familiar American, European and Chinese models suggest. The debate about global AI governance is often organised around three reference points : America as an innovation-led model, the European Union as a regulation-led model and China as a state-led model. But these labels do not describe the same kind of thing. “American innovation” points more to what AI should deliver. “European regulation” points to how AI should be governed. And “Chinese state-led development” combines both a governance style and a development agenda. This matters because these three models are not just descriptions of America, the EU and China. They also tend to serve as a map : other countries are placed closer to one pole or another, as though every national approach were a variant of the American, European or Chinese model. But countries are not simply choosing between three ready-made options. They are combining different answers to two separate questions: how should AI be governed and what should AI be used to achieve? Once we separate those dimensions, the familiar three-model story becomes much less convincing. These choices are visible in national AI strategies , the official documents in which governments set out their priorities for AI development, adoption and governance. Some give more weight to ethics, regulation and trust. Others emphasise growth, innovation, public-sector transformation, data infrastructure, workforce development or climate. Analysing these documents systematically reveals a more varied global landscape than the three-model shorthand suggests. Our approach We analysed the 56 countries that currently have a flagship national AI strategy catalogued on x.Machina , the open-access governance mapping platform we co-founded. For each, we took the most recent strategy document and scored it across eight themes: ethics, regulation, growth, innovation, public services, data, workforce and climate. The scoring is text-based rather than keyword-based, so two countries can be close in the analysis even when they use different vocabulary. From those eight themes we built two composite measures. Governance orientation measures the balance between ethics-and-regulation language and growth-and-innovation language. Functional emphasis captures how much the text stresses delivery-oriented themes, including growth, innovation, public services and data infrastructure. The result is shown in Figure 1, with four archetypes emerging from the clustering. Figure 1: National AI strategy typology (n=56). Horizontal axis: governance orientation. Vertical axis: functional emphasis. Colours: archetypes from clustering on full thematic profiles. The anchors don’t anchor Neither America, the EU or China anchor a single cluster in the way that the three-model story implies. America does not anchor an Anglosphere-led innovation grouping. If it did, the cluster around it would contain its closest political and economic peers. It contains some of them, including Australia, Britain and Ireland. But it also contains Bangladesh, Brazil, Colombia, Ethiopia and Moldova, middle- and lower-income economies the bloc story would never put in an American-led camp. The European Union does not anchor a single cluster. France and Germany appear in the largest cluster, Balanced (Ethics + Regulation) , alongside Canada, Malaysia, Mexico, New Zealand, Saudi Arabia, South Korea and several smaller jurisdictions. But Austria, Belgium, Denmark, Italy and Sweden sit in a different cluster, Growth-oriented (Climate + Data) , the same cluster as America, which is characterised by stronger emphasis on delivery themes. The EU member states are not all writing the same kind of strategy, and several non-EU jurisdictions are closer to some EU members than other EU members are. No Chinese-model bloc emerges from the texts. The Balanced (Economic + Public) cluster groups China with Argentina, Estonia, Greece, India, Japan, Lithuania, Peru and Singapore. India and Japan would never be described as following a Chinese model. Yet their AI strategies share China’s combination of priorities: a weighting of economic and public-sector themes without a strong ethics-and-regulation lean. The “China bloc” of geopolitical commentary has no equivalent in the text of national AI strategies. A small fourth cluster, Low-salience (Ethics + Regulation) , groups Cyprus, Finland, Jamaica, Malawi and the United Arab Emirates, whose strategies engage only lightly with each of the eight themes. Why the bloc story misleads What emerges from the analysis is a set of approaches richer than the three models suggest. Each cluster reflects a particular combination of priorities, in how a country talks about governing AI and what it wants AI to deliver – and the combinations rarely map onto the familiar American, European or Chinese categories. The countries that share each combination are a varied mix: politically, economically and regionally. Germany and Saudi Arabia, for instance, share a cluster; America sits alongside Bangladesh and Brazil; China alongside India and Japan. Reading along these two dimensions also surfaces unexpected similarities. China and Scotland , whose recent AI strategy is also included in the x.Machina dataset, sit in different archetypes and differ on governance orientation, which is roughly what a bloc-map reader would expect. But on functional emphasis (how much each text engages delivery themes like infrastructure, public services and innovation deployment), the two are strikingly close. Two countries cast as opposites in the bloc story turn out to weight delivery themes almost identically. The implication for anyone working across jurisdictions is concrete. The three-model story is a poor guide to what a country’s AI strategy actually emphasises. It will systematically mislead anyone using it to anticipate regulatory direction, public-sector AI priorities or where institutional architectures are likely to converge. Reading the texts themselves, along the two dimensions set out here, is a more reliable basis for cross-jurisdictional analysis. Similarities like the one between China and Scotland sit inside the archetypes rather than between them, and would need closer comparison than the four archetypes alone can offer. The picture can change over time The story becomes more complicated still when we consider that countries often produce more than one AI policy document. The American strategy used in this analysis is the 2025 Action Plan, a comprehensive policy document that places the country in the Growth-oriented cluster. The March 2026 legislative recommendations, a newer, thinner and more targeted legislative text, would tell a different story and place it elsewhere on the map. Both are formal United States government AI policy documents, but they perform different functions. This is not a problem for the analysis; it is part of the point. National AI strategies are not fixed expressions of national identity. They are policy texts written for particular purposes, at particular moments. Where a country sits on the map depends partly on which document is treated as representative. What this analysis reveals is less a static map than a moving picture: a global landscape continuously rewritten as governments produce, revise, and supersede their AI policies. The full x.Machina dataset and the analysis behind this article are publicly available at: https://www.nomosmachina.org/ This article gives the views of the author, not the position of LSE Business Review or the London School of Economics. You are agreeing with our comment policy when you leave a comment. Image credit: K illustrator Photo provided by Shutterstock. The post How countries write their AI futures – mapping the many models of AI governance first appeared on LSE Business Review .

Summary generated from the RSS feed of LSE Business Review. All article rights belong to the original publisher. Click through to read the full piece on blogs.lse.ac.uk/businessreview.