San Francisco uses AI to streamline municipal codes

The San Francisco City Attorney's Office announced it is using artificial intelligence to analyze its municipal codes--a development that could eventually be applied to local ordinances and federal laws, according to a law professor involved in the project.

City Attorney David Chiu's office said Thursday it had partnered with Stanford RegLab to deploy an AI tool that analyzes statutes and flags opportunities for reform. The tool, named STARA, analyzed the entirety of the city's municipal codes and resolutions, totaling nearly 16 million words, according to a news release.

Chiu introduced an ordinance proposing the elimination or consolidation of 174 public reports--roughly 36% of amendable reporting requirements--across city departments, the release stated.

Stanford Law School professor and director of the RegLab, Daniel Ho, said in a phone interview Thursday that the tool has "tremendous potential" to reshape the approach to bureaucratic processes in government agencies.

"It's a pioneering use case of relying on AI to help sift through mountains of code that would be too consuming otherwise to tackle for code reform and simplification," said Ho.

Chiu's proposed ordinance is to be considered by the San Francisco Board of Supervisors in the coming months.

The possibilities of the tool's application to identify forms of "policy sludge" that have "weighed government down" could also be used to address outdated laws still in use, according to Ho.

"One can use this kind of a tool to do what Ruth Bader Ginsburg did when she was the Columbia law professor, which is she hired an army of research assistants to spend months going through the U.S Code, for instance, to identify all examples of gender discrimination that existed within the U.S Code," Ho said.

STARA can "radically reduce the cost" of this kind of search, which makes it different from an "off the shelf AI chatbot," Ho said.

The tool has been trained to read the code and implement a "statutory tree that preserves statutory definitions, cross-references to other sections, and importantly, the structure of the code, which is really important, obviously, for how lawyers tend to think of the text and context," Ho explained.

The City Attorney's Office then consulted with city departments to determine which reports were obsolete, duplicative or unnecessary, according to the news release. One example provided by the city was the production of a biennial report on "fixed pedestal zones" for newspaper racks, which no longer exist.

Ho said the tool was tested on the U.S. Code before using it on San Francisco's municipal codes. The news release stated that at the federal level, the number of mandated reports has been described as a congressional "black hole."

"This collaboration is a great example of how governments can use AI to make impactful change," Chiu said in the media release. "RegLab's tool saved us countless hours of work. ... Deleting obsolete requirements will free up staff time to focus on core service delivery, particularly during challenging budget times."

While other cities across California have reported using a variety of AI tools to increase efficiency, Ho said it would be fair to say San Francisco appears to be the first to use it to shape legislation.

The city of San Jose has also committed to using AI to draft documents and assist with city services, according to its AI guidelines. It also founded the GovAI coalition in November 2023 "to promote responsible and purposeful AI in the public sector" with more than 600 member agencies across the country.

james_twomey@dailyjournal.com