Blab AI
Stanford AI Tool Reveals Decades-Long Bloat in U.S. State Statutes, Prompting New York and San Francisco Reforms
Stanford researchers have used an artificial‑intelligence system to scan 500 million words of state statutes, uncovering widespread “policy sludge” that has accumulated over decades. The study, led by the Stanford Institute for Human‑Centered AI (HAI) and the Regulation, Evaluation, and Governance Lab (RegLab), found that California’s reporting requirements grew 400 % from 2000 to 2025, with many reports never completed or filed. The findings are already influencing policy: New York Governor Kathy Hochul issued an executive order this month directing a state‑wide regulatory reset, and San Francisco has passed legislation that streamlines more than a third of its reporting rules.
Policy sludge refers to obsolete provisions, redundant reporting obligations, and other bureaucratic frictions that clog public‑sector legislation. The RegLab team built a domain‑specific AI system—named STARA—to perform comprehensive statutory surveys. By feeding the system the full corpus of state statutes, the researchers were able to identify reporting requirements, commissions, and fees across all 50 states. The tool was first validated in partnership with the San Francisco City Attorney in 2023; the resulting analysis led to legislation that reduced the city’s reporting burden.
The cross‑state analysis produced several striking findings. First, reporting requirements have ballooned over time. California’s obligations grew 400 % between 2000 and 2025, while in Maryland the time required to read all reports could reach 14 weeks—longer than a typical legislative session. Second, many reports are never filed; in California, 30 % of ongoing reports may never be completed, and Maryland agencies identified 20 % of reports as candidates for elimination or consolidation. Third, the cost of producing a single report can vary dramatically: one study estimated a report required 3,500 staff hours and more than $870,000 in resources, whereas other reports cost only a few hours and receive thousands of views. Finally, the team found that while reporting requirements are somewhat more common in Democratic states, the partisan difference is small relative to the overall scale of the problem.
The RegLab team has worked directly with state governments to translate these findings into action. In New York, the executive order directs agencies to remove outdated requirements, burdensome fees, and unnecessary reports and commissions. Zoe Jacobs, director of regulatory reform in Governor Hochul’s office, said the AI tool helped convert legalese into digestible datasets that enable staff to systematically review mandated reports. In California, the state is using the data to convert paper reports into digital dashboards where possible. Maryland’s analysis included an estimate of the benefits and costs of reporting requirements across four agencies.
Beyond specific reforms, RegLab is providing a model state statute that includes automatic sunsetting, a digital repository, and lightweight tracking of costs and benefits. The team also released the full scan results and a public website that allows anyone to explore reporting requirements across the country. The forthcoming paper, “The Abundance of Reports and Incapacity of States,” will appear in the Yale Journal on Regulation.
The work illustrates how AI can support evidence‑based government reform. By identifying obsolete provisions and quantifying the burden of reporting, the RegLab tool offers a practical pathway for states and cities to streamline their statutes. As more jurisdictions adopt similar approaches, the policy‑sludge problem may shrink, improving the efficiency of public administration and reducing the administrative burden on both civil servants and citizens.
In the coming months, the RegLab team will continue to refine its AI system and expand its collaborations. The public resource website will be updated with new data as additional states conduct their own analyses. Meanwhile, New York’s regulatory reset is expected to roll out in phases, and San Francisco’s streamlined reporting rules will take effect in the next legislative session. The broader AI‑in‑government community will likely monitor these developments closely as a model for data‑driven regulatory reform across the United States.
Policy sludge refers to obsolete provisions, redundant reporting obligations, and other bureaucratic frictions that clog public‑sector legislation. The RegLab team built a domain‑specific AI system—named STARA—to perform comprehensive statutory surveys. By feeding the system the full corpus of state statutes, the researchers were able to identify reporting requirements, commissions, and fees across all 50 states. The tool was first validated in partnership with the San Francisco City Attorney in 2023; the resulting analysis led to legislation that reduced the city’s reporting burden.
The cross‑state analysis produced several striking findings. First, reporting requirements have ballooned over time. California’s obligations grew 400 % between 2000 and 2025, while in Maryland the time required to read all reports could reach 14 weeks—longer than a typical legislative session. Second, many reports are never filed; in California, 30 % of ongoing reports may never be completed, and Maryland agencies identified 20 % of reports as candidates for elimination or consolidation. Third, the cost of producing a single report can vary dramatically: one study estimated a report required 3,500 staff hours and more than $870,000 in resources, whereas other reports cost only a few hours and receive thousands of views. Finally, the team found that while reporting requirements are somewhat more common in Democratic states, the partisan difference is small relative to the overall scale of the problem.
The RegLab team has worked directly with state governments to translate these findings into action. In New York, the executive order directs agencies to remove outdated requirements, burdensome fees, and unnecessary reports and commissions. Zoe Jacobs, director of regulatory reform in Governor Hochul’s office, said the AI tool helped convert legalese into digestible datasets that enable staff to systematically review mandated reports. In California, the state is using the data to convert paper reports into digital dashboards where possible. Maryland’s analysis included an estimate of the benefits and costs of reporting requirements across four agencies.
Beyond specific reforms, RegLab is providing a model state statute that includes automatic sunsetting, a digital repository, and lightweight tracking of costs and benefits. The team also released the full scan results and a public website that allows anyone to explore reporting requirements across the country. The forthcoming paper, “The Abundance of Reports and Incapacity of States,” will appear in the Yale Journal on Regulation.
The work illustrates how AI can support evidence‑based government reform. By identifying obsolete provisions and quantifying the burden of reporting, the RegLab tool offers a practical pathway for states and cities to streamline their statutes. As more jurisdictions adopt similar approaches, the policy‑sludge problem may shrink, improving the efficiency of public administration and reducing the administrative burden on both civil servants and citizens.
In the coming months, the RegLab team will continue to refine its AI system and expand its collaborations. The public resource website will be updated with new data as additional states conduct their own analyses. Meanwhile, New York’s regulatory reset is expected to roll out in phases, and San Francisco’s streamlined reporting rules will take effect in the next legislative session. The broader AI‑in‑government community will likely monitor these developments closely as a model for data‑driven regulatory reform across the United States.
