Global AI rules don't really work; countries only pass tough laws based on their own internal history, not global peer pressure.
A massive analysis of 8,000 documents found that binding AI laws are not driven by international trends or peer-country influence. Instead, they are the result of a slow, domestic 'law hardening' process where a country piles up strategy memos until they eventually turn into binding legislation, debunking the idea of global regulatory diffusion.
<p>AI Policymaking Across the Globe: A Large-N Analysis of More Than 8,000 Policy Documents from 76 Countries</p>
SSRN · 6312103
Legislative activity on AI governance has surged globally, yet we lack comparative empirical evidence on how "law hardening" develops and which pre-legislative activities catalyse binding regulation. We analyse the AI policy formation pipeline using more than 8,000 policy documents from 76 countries. Using Cox survival models, event studies, and Granger-style tests, we examine whether binding AI laws are driven by domestic pre-legislative activity or cross-border diffusion. We find that domestic