The Bottleneck in AI Governance: Evidence from 1,419 State Bills

Understanding what makes AI governance succeed or stall has immediate practical consequences for society. Since 2017, U.S. state AI bill volume has grown more than 35-fold, yet most never reach a floor vote. We present the first state-level bill outcome analysis to decompose the AI governance bottleneck into two distinct gates, analyzing 1,419 bills across all 50 states from 2017 to 2026. Our two-stage hurdle model reveals the bottleneck is not monolithic: specialist committee referral is the primary barrier to a floor vote, while cross-party alignment is the decisive factor at enactment. States are enacting AI laws at $8\times$ the federal rate, making them the viable near-term venue for AI governance. For the ML and AI community, our findings translate directly into where and how the AI community should engage in the legislative processes to advance trustworthy AI governance.