From Ambition to Action: National AI Strategies and Downstream Policymaking

• In Event: Accountable Government

Abstract

National AI strategies have become a common instrument of AI governance, yet governments often make AI policy before adopting them. This paper asks whether formal strategy adoption changes subsequent policymaking, where post-adoption policy growth occurs, and whether later policy follows the priorities stated in the strategy. I construct a 74-country panel from a frozen snapshot of the OECD.AI Policy Navigator and exploit variation in the timing of first national AI strategy adoption using a staggered difference-in-differences design. I combine this analysis with domain-level policy outcomes, LLM-assisted coding of first-strategy priorities, and the V-Dem Liberal Democracy Index. Strategy adoption is followed by approximately 2.17 additional substantive AI policy initiatives per country-year over the first five full post-adoption years. This growth follows an implementation hierarchy: it is strongest in enabling and governance-capacity domains, followed by operational safeguards, and weakest in legal and rights-based constraints. However, prioritized domains do not experience significantly greater post-adoption growth once country-wide policy expansion and global domain-specific trends are absorbed. Liberal democracy likewise does not systematically predict stronger rights-oriented prioritization or strategy-to-policy translation. These findings show that national AI strategies appear to mobilize policymaking more effectively than they direct its substantive allocation. They also demonstrate the importance of distinguishing policy mobilization, the composition of policy growth, and strategy-to-policy matching when evaluating national AI strategies.

Author

• Jing Tao, University of Texas, Dallas
  Presenting Author