The Class of 2019: What Triggered the First Wave of National AI Strategies

In 2017, three countries published national AI strategies. By the end of 2019, that number had grown to 43. What happened?

The AI Folio corpus documents the full chronology across 106 documents from 87 jurisdictions. The pattern is unambiguous: 2019 was the pivotal year, with 20 new national strategies published — more than in the preceding two years combined, and a figure that would not be matched again until 2024.

Three forces converged to produce the 2019 wave. First, the publication of the OECD AI Principles in May 2019 gave governments a legitimate, internationally endorsed framework to reference — and the corpus reflects this directly, with OECD cited in 59 of 104 strategies. Countries that had been watching and waiting suddenly had a template. Second, competitive signalling pressure from early movers had become impossible to ignore. Canada's 2017 Pan-Canadian AI Strategy and France's 2018 Villani Report had established that national AI strategies were now expected signals of technological seriousness. Third, 2019 saw AI feature as a substantive policy issue in elections across multiple jurisdictions, creating domestic political incentives to publish regardless of whether strategies had substantive content.

After 2019, adoption continued at a more measured pace. The 2020-2021 period produced 24 new strategies, many accelerated by the experience of deploying AI tools for pandemic response — contact tracing, resource allocation, epidemiological modelling. The urgency of that deployment forced governments that had been developing strategies slowly to publish sooner.

The most striking feature of the full timeline is the second acceleration in 2024-2025: 34 new strategies in two years, matching the 2019 wave in scale. The trigger here is less ambiguous. The EU AI Act came into force in August 2024, creating direct compliance obligations across the bloc and a powerful demonstration effect for non-EU countries — particularly in the Western Balkans, North Africa, and parts of Asia — that monitor European regulatory developments closely.

The ChatGPT launch in November 2022 produced surprisingly few immediate new strategies. The policy response to generative AI has been slower than the public response. Most governments chose to revise or supplement existing strategies rather than publish new ones — a pattern that means the corpus almost certainly undercounts the genuine governance response to the generative AI moment. The strategies are there; they are just updates, not new documents.

The two waves — 2019 and 2024-2025 — were triggered by different mechanisms. The first by normative entrepreneurship and competitive signalling; the second by hard regulatory pressure. That difference matters for what the strategies say and how seriously they should be taken.

Chart: Bar chart — strategies published per year 2017-2026. Annotated with: OECD Principles (May 2019), ChatGPT launch (Nov 2022), EU AI Act in force (Aug 2024).

Data: AI Folio Corpus, year distribution of 106 strategy documents across 87 jurisdictions.