Policy Convergence and Divergence Across National and Within Regional AI Strategies: A Policy Design Element Analysis

Governments worldwide have responded to the rapid expansion of AI by publishing national and regional AI strategies. Comparing national and regional AI strategies to identify their convergences and divergences can uncover their common practices, understand regional variations, and provide policy designers a comprehensive set of policy design elements for their ongoing AI strategy developments. Yet, existing work has not examined their underlying policy design elements or assessed whether those elements are horizontally (country-to-country) or vertical (region-to-country) converging or diverging over time. This paper addresses that gap by coding and analyzing 74 national and 3 regional AI strategies drawn from a global scan of all 205 UN member and non-member states. The coding used a latent-inductive approach organized around three functional policy design elements: goals, approaches, and principles. Two research questions guided the analysis: to what degree are national AI strategies becoming horizontally convergent or divergent over time; and to what degree are national strategies becoming vertically convergent or divergent with those countries’ regional AI strategy. Results indicate strong horizontal convergence around economic competitiveness, research support, and ethical AI use, alongside persistent divergence in human rights goals, participatory governance approaches, and human-centric principles. Across the three regions, the AU exhibits the highest vertical convergence, the EU demonstrated strong alignment on regulatory and economic priorities but diverges on human-centric values, and the Nordic-Baltic Region displays mixed vertical convergence. These findings offer policy designers a comprehensive evidence base for identifying emerging AI policy design choice norms as AI strategies are developed and updated.

The study adopts a policy design approach – where policy design refers to the deliberate selection and combination of elements within a policy instrument, particularly its-goals, -approaches, and-principles. Policy-goals define what a policy aims to accomplish; policy-approaches specify the mechanisms through which those goals are to be pursued; and policy-principles articulate the normative values and standards that should be upheld throughout the goal-attainment process. Policy convergence is defined as the tendency for policies across distinct jurisdictions to be or become more alike over time in their structures and substantive commitments, a process that can occur horizontally (across jurisdictions at the same administrative level), or vertically (between different administrative levels within the same hierarchy, such as a regional body and its Member states). Importantly, convergence and divergence can coexist within the same domain, producing divergence-within-convergence, where jurisdictions adopt similar high-level governance language while diverging on the specific approaches and principles used to operationalize it.

The results reveal a landscape characterized by strong convergence around certain economic and institutional priorities alongside persistent divergence in substantive social commitments. Across national strategies, near-complete convergence is observed around supporting national AI research (71/74 strategies) and ethical AI use (68/74 strategies), suggesting these elements have effectively become baseline expectations for any credible national AI strategy. At the same time, divergence persists around human-rights goals, participatory governance approaches, and human-centric principles like diversity and inclusion, pointing to a structural gap between commitments and the substantive choices that would operationalize those commitments. Within regions, the AU exhibits the highest vertical convergence, the EU demonstrates strong alignment on regulatory and economic goals but diverges on human-centric principles, and the NBR displays mixed vertical convergence with alignment around broad ethical values but divergence across operational goals and approaches. Together, these findings suggest that the window for independent AI policy problem-solving is narrowing, that ethics washing risks eroding public trust in strategies that adopt ethical language without substantive follow-through, and that regional strategies that are too abstract risk producing minimal national convergence, all of which has implications for how policy designers approach the development and updating of national and regional AI strategies.

Horizontal patterns show a general trend towards convergence of policy-goals over time. These data suggest a strong convergence around economic and capacity-building properties. Build Capacity (61/74 strategies) and Become an AI Leader (59/74 strategies) dominate the landscape, remaining consistently present across all years. This persistent emphasis indicates that countries are aligning with the view that national competitiveness and capacity-building are core AI policy-goals. Specific policy-goal codes also reveal areas of convergence. Through Research and Development appears in 56, while Mentions Innovation and Through Skills each appear in 49 strategies. These consistent inclusions point to a large consensus that investing in knowledge and skills is essential to remain competitive in AI. However, this convergence is not universal. Human rights (23/74 strategies) and national security (15/74 strategies) appear sporadically, suggesting divergence in how countries integrate broader social or security concerns into their AI policy-goals.

For policy-approaches, the most striking trend is the near-dataset-complete adoption of Support National AI Research, which appears in 71/74 strategies. This broad code, along with its specific-codes (research, collaboration, and knowledge transfer) has steadily increased, suggesting that countries are converging on research and innovation as the basis of their AI policy-approach. Other widely shared approaches, such as Public Sector AI Adoption and Automate Processes, Increase Efficiency and Transparency, also demonstrate convergence. At the same time, divergence remains. Less frequently adopted policy-approaches, like Encourage Public Consultations or Focused on Legislative and Regulatory Efforts appear in fewer than 30 strategies. This limited uptake indicates weaker convergence around participatory and regulatory pathways, suggesting that while countries align on research and efficiency-oriented strategies, they diverge in how they involve citizens or pursue regulatory frameworks.

For policy-principles, the principle of Ethical AI Use dominates, appearing in 68/74 strategies, demonstrating considerable horizontal convergence around the importance of ethical AI use. Specific codes, like privacy, transparency, and trust are consistently present, showing that countries are converging around a shared set of ethical guidelines. Beyond ethics, principles like Explainability and Reliability show notable increases after 2020, suggesting that policy convergence is extending beyond broad ethics to technical governance principles. However, divergence appears in principles tied to human values and inclusion. For example, Respects Human Dignity and Ensures Human Diversity and Inclusion appears in fewer than 35 strategies and are unevenly distributed over time. Similarly, Collective Debate and Use Media are rarely mentioned. These absences reveal divergence in how far countries go in embedding societal inclusion and participatory governance into their AI strategies.

Turning to vertical patterns, the EU exhibits a mixed picture of vertical convergence. In terms of policy-goals, strong convergence is evident around economic ambition and capacity-building. Both the European Commission (EC) and member states emphasize becoming global AI leaders and building national capacity. However, divergence emerges in policy-goals related to human rights and national security, which member states mention more often than the EC. Policy-approaches reflect similar dynamics. EU Member states align closely with the EC on building human capacity and expanding AI-related skills. Yet, divergence appears in areas like diversity and inclusion, which the EC prioritizes but many states underemphasize. Policy-principles show the greatest vertical convergence as the EC and Member states emphasize regulatory frameworks and ethical governance. Still, divergence exists in policy-principles around diversity and inclusion, which remain less prominent at the national level. Overall, the EU demonstrates strong vertical convergence on regulatory and economic goals but divergence around human-centric values.

The AU shows strong vertical convergence across policy-goals, -approaches, and-principles. Most AU-endorsed policy-goals, like stimulating growth and solving societal challenges are echoed by its Member states. Minimal divergences were found, with just two national goals absent from the AU framework. Policy-approaches reinforce this pattern. National to regional strategies converge on education, skills, and research as central pillars. Only a handful of distinct national policy-approaches diverge from the AU's regional strategy, indicating strong convergence. Policy-principles also strongly convergence in their emphasis of ethical use and data protection. Divergence appears only where some Member states introduce forward-looking principles like attracting international AI talent, which the AU has not yet codified. Overall, the AU region demonstrates the highest vertical convergence among the three regional-national strategies.

The NBR shows mixed vertical convergence. Policy-goals reveal five shared priorities between regional and national levels. While both emphasize ethics and democracy, many national goals are absent at the regional level, producing mixed convergence. Policy-approaches are similarly mixed. Only ten broad and specific policy-approaches are shared between the regional and various national AI strategies, and while some NBR states emphasize inclusive governance, others focus on innovation and entrepreneurship, which are not consistently reflected in the NBR AI strategy. This mix suggests a lack of harmonized vision compared to the EU and AU cases. Policy-principles show modest convergence around transparency and ethics, but divergence elsewhere. National strategies often expand on principles not endorsed at the regional level, like talent attraction and research publication. As a result, vertical convergence in the NBR is mixed, with convergence in core ethical concerns but divergence in broader priorities.

The utility of these findings for policy designers is fourfold. First, systematic policy convergence analysis enables policy designers to engage in what Bennett (1991) called lesson-drawing – the deliberate and structured analysis of first-mover states' policies to design better and more informed domestic policies. Second, understanding where convergence is and is not occurring allows policy designers to identify emerging norms that may harden into assumed policy design standards. Third, tracking temporal convergence and divergence across all three policy design element types helps policy designers avoid the trap of policy layering, drift, and incoherence. Fourth, vertical convergence analysis gives policy designers working within regional governance frameworks, whether EU, AU, or NBR Member states a precise understanding of the degree to which their national strategies converge or diverge from the regional strategy.

Improvements to AI Systems Based on the Paper:

1. AI-Powered Policy Convergence Analyzer

• Automatically codes and compares AI strategies (national/regional) across goals, approaches, and principles using latent-inductive NLP.

• Detects horizontal (country-to-country) and vertical (region-to-country) convergence/divergence over time, flagging emerging norms and gaps (e.g., ethics washing).

• Outputs a live dashboard for policymakers showing which elements are becoming standardized vs. fragmented, enabling real-time lesson-drawing from first-mover states.

2. Ethics Washing Detection System

• Scans AI strategies for high-level ethical language (e.g., ethical AI use) and cross-checks against concrete operational approaches (e.g., participatory governance, human rights goals, diversity principles).

• Flags strategies that adopt ethical rhetoric without substantive follow-through, quantifying the divergence-within-convergence pattern identified in the paper.

• Provides a risk score for public trust erosion, helping governments align stated principles with actionable policies.

3. Regional Alignment Advisor

• For a given country, compares its draft AI strategy against its regional framework (e.g., EU, AU, NBR) in real time.

• Identifies specific policy-goals, -approaches, and-principles where the country diverges vertically (e.g., missing human-centric principles in EU states, or overemphasis on innovation in NBR states).

• Recommends adjustments to increase vertical convergence, based on the paper’s finding that AU shows highest convergence and NBR shows mixed alignment.

4. Temporal Policy Drift Monitor

• Tracks changes in AI strategies over time to detect policy layering, drift, or incoherence (e.g., a country adding human rights goals but not updating approaches or principles).

• Uses the paper’s coding framework to alert designers when new goals are introduced without corresponding operational mechanisms, preventing internal contradictions.

• Benchmarks a country’s evolution against global trends (e.g., post-2020 rise in explainability and reliability principles).

5. AI Strategy Gap-Filling Generator

• Given a country’s current AI strategy, identifies missing high-convergence elements (e.g., support national AI research appears in 71/74 strategies) and suggests context-specific additions.

• For low-convergence areas (e.g., human rights goals, participatory governance), generates optional policy language and mechanisms, allowing designers to choose whether to align with emerging norms or intentionally diverge.

• Provides evidence-based rationale for each suggestion, citing convergence rates and regional patterns from the paper.

6. Cross-Regional Benchmarking Tool

• Compares a user-selected set of national AI strategies against each other and against regional frameworks, visualizing convergence clusters (e.g., economic goals vs. human-centric principles).

• Highlights where a country is an outlier (e.g., only 15/74 mention national security) and predicts potential future convergence pressures based on temporal trends.

• Enables policy designers to anticipate which elements may harden into standards, using the paper’s finding that economic and research goals are near-universal.

Governments worldwide have responded to the rapid expansion of AI by publishing national and regional AI strategies. Comparing national and regional AI strategies to identify their convergences and divergences can uncover their common practices, understand regional variations, and provide policy designers a comprehensive set of policy design elements for their ongoing AI strategy developments. Yet, existing work has not examined their underlying policy design elements or assessed whether those elements are horizontally (country-to-country) or vertical (region-to-country) converging or diverging over time. This paper addresses that gap by coding and analyzing 74 national and 3 regional AI strategies drawn from a global scan of all 205 UN member and non-member states. The coding used a latent-inductive approach organized around three functional policy design elements: goals, approaches, and principles. Two research questions guided the analysis: to what degree are national AI strategies becoming horizontally convergent or divergent over time; and to what degree are national strategies becoming vertically convergent or divergent with those countries' regional AI strategy. Results indicate strong horizontal convergence around economic competitiveness, research support, and ethical AI use, alongside persistent divergence in human rights goals, participatory governance approaches, and human-centric principles. Across the three regions, the AU exhibits the highest vertical convergence, the EU demonstrated strong alignment on regulatory and economic priorities but diverges on human-centric values, and the Nordic-Baltic Region displays mixed vertical convergence. These findings offer policy designers a comprehensive evidence base for identifying emerging AI policy design choice norms as AI strategies are developed and updated.