Regulator and User: A Comparative Analysis of State AI Policies for Government Operations

• In Event: AI in Public Procurement, Budgeting, and Government Operations

Abstract

Artificial intelligence is increasingly used inside government agencies for decision-making, service delivery, and daily operations. As AI becomes more common in the public sector, governments need clear rules about how to use it responsibly. Regulating the use of AI is a rather contentious issue, however, especially given the Trump administration’s attempted moratorium aimed at prohibiting state AI policies. Regardless, many states in the US have issued their own AI policies, guidelines, and executive orders to govern how AI is used in government. Existing research has examined state AI legislation broadly (DePaula et al., 2024, 2025) and how state policies address accountability specifically (Chen et al., 2026). However, a unique and underexplored challenge remains: state governments act as regulators while both state agencies and local governments serve as the users subject to those rules. We know little about how state governments have approached this dual role, specifically, what rules they have created for their own agencies and employees, and how those rules differ across states. This study addresses that gap. Drawing on the AI governance and public accountability literature (Wirtz et al., 2019; Harrison & Luna-Reyes, 2022), we treat state AI policy documents as formal expressions of how governments understand their responsibilities when using AI. This represents an intragovernmental regulation, where one body of the government sets rules that other government bodies follow (Waterman & Meier, 1998). This lens allows us to move beyond simply cataloging existing policies toward understanding the governance logic embedded in them.

This study asks two research questions. First, how do state governments regulate, guide, or govern the use of AI in state and local agencies? Second, what differences exist across state policies in terms of the scope of AI covered, regulatory stance, and accountability mechanisms?

To answer these research questions, we collected around 70 policy documents across 47 states in the United States. These documents consist of executive orders, state legislation, guidelines, frameworks, etc., that aim to govern how AI is to be used by public agencies in their daily operation. We will utilize qualitative content analysis as it is appropriate for systematically examining policy documents. We developed a coding scheme to analyze these documents along five dimensions: 1) what types of AI the policy covers; 2) the regulatory stance (whether AI use is allowed by default or must be approved first); 3) who is responsible for take responsibility of AI usage 4); whether and how human oversight is required; 5) how mature and comprehensive the policy is. By comparing different state policies, we aim to observe differences between states in these five dimensions and provide multidimensional comparisons of how intragovernmental regulation is framed and embedded in the AI governance practices. This research contributes to emerging scholarship on public sector AI use, AI governance, and AI accountability, while offering practical benchmarks for states at earlier stages of policy development. At a moment when federal AI oversight remains uncertain, understanding how states govern the use of AI is especially consequential.

Author

• Md Abu Talha, University at Albany
  Presenting Author