The new 'diffusion' models? Tracing AI policy diffusion across all 50 states

• In Event: Politics and Policymaking in State-Level AI Governance

Abstract

As the use of AI across different sectors has grown significantly in the past 10 years, there has been an increasing awareness of not only the benefits presented by these technologies, but also risks such as violations of privacy, deepfakes, racial bias, and many other concerns. In the absence of Federal AI regulation in the US, AI-related laws have been introduced in all 50 states (National Conference of State Legislatures 2025). However, few studies have examined these bills in depth. The limited existing scholarship on AI state legislations illustrates the strong influence of big tech actors and advocacy groups on state laws, as well as the diffusion of legislations across different states (Hinkle, 2022). Thus, in this paper, I conduct policy document analysis on a few different datasets detailing state-level AI bills introduced and enacted over the past eight years. Using Latent Dirichlet Allocation (LDA), I uncover common themes across the different state legislations, and identify relationships and networks across them. In addition, I conduct time series analysis to understand how AI policies diffuse across different states. Applying policy diffusion theory and diffusion methods, and building on previous scholarship on state-level policy diffusion (DellaVigna and Kim 2025), I undertake statistical analysis to understand whether AI policy diffusion at the state level is impacted by internal factors, such as party alignment and proximity, or external factors, such as the influence of advocacy groups. To that end, I specifically consider the impact of unprecedented tech lobbying efforts on both bill introduction and policy diffusion across states. To better understand the context of state lobbying efforts, I analyse trends in publicly available state lobbying data. Overall, the paper contributes to existing scholarship on both AI policy and policy diffusion, and has implications for understanding the mechanics of state-level introduction and adoption of AI-related bills going forward. This remains important at this early stage of AI policy adoption, as decisions made at this stage are likely to impact the trajectory of policy adoption and diffusion going forward.

Author

• Bhargavi Ganesh, University of Pennsylvania
  Presenting Author