The ISO Surprise: Rethinking Who Shapes Global AI Governance Norms

When governments write AI strategies, they do not write in isolation. They reference international frameworks, cite peer countries, and signal alignment with global norms. The pattern of these citations reveals the invisible architecture of global AI governance — and the results contain at least one significant surprise.

Across 104 strategy documents, we tracked explicit mentions of international frameworks and forums. The conventional assumption is that the OECD AI Principles dominate — and OECD does appear frequently, cited in 59 strategies. But the most cited framework in the entire corpus is ISO, referenced in 78 strategies. More countries mention ISO standards than any other international body when writing their national AI governance documents.

This finding reframes the conventional narrative about AI governance norm-setting. Academic and policy attention has focused heavily on the OECD Principles, UNESCO's Recommendation, and the EU AI Act as the primary vectors of normative influence. ISO's dominance in citation frequency suggests that technical standardisation bodies may be shaping national AI governance more quietly but more pervasively than the high-profile political frameworks.

NATO appears in 42 strategies — a figure that reflects both the security dimension of AI governance and the breadth of NATO's membership and partnership networks. UNESCO's Recommendation on the Ethics of AI appears in 35 strategies, predominantly among countries without strong OECD ties. For many Global South countries, UNESCO represents a more legitimate multilateral voice than an organisation whose membership they do not hold.

The UN appears in 30 strategies, G20 in 22. The Bletchley Declaration (2023), despite significant diplomatic attention at the time, leaves a modest footprint in the corpus — appearing in relatively few strategies and almost exclusively among summit attendees. This suggests its influence has been more diplomatic than normative: a signal of relationship rather than a durable source of governance ideas.

The citation hierarchy matters because it tells us whose definitions, whose principles, and whose framings are shaping how governments think about AI. A corpus where ISO dominates is a corpus where technical interoperability and standards compliance are the primary governance frame — not ethics, not rights, not democratic accountability. Whether that is the right frame for AI governance is the central question that the documents themselves tend not to ask.

Chart: Horizontal bar chart — framework citation counts: ISO (78), OECD (59), NATO (42), UNESCO (35), UN (30), G20 (22), and others. Sorted descending.

Data: AI Folio Corpus Metrics, international alignment analysis of 104 national AI strategy documents.