San Diego explores AI tool to draft city policies and staff reports

San Diego is in active discussions with Madison AI about deploying artificial intelligence to draft city policies, staff reports, and legislative documents by ingesting the municipality's entire document archive.

The city has been evaluating the AI platform since at least December, when Madison AI representatives met with city staff and presented their system. Public records obtained by Axios San Diego show the company's model would consume "every single PDF and archived file" containing the city's institutional knowledge to generate new policy text.

Why it matters

This represents a significant expansion of AI use in municipal governance beyond administrative tasks, with the technology potentially authoring the actual policy language that elected officials vote on. The approach raises fundamental questions about data security practices and whether algorithmic systems should replace human deliberation in democratic processes.

How the system works

Madison AI's platform analyzes a city's complete document repository to produce legislative materials. According to the company's presentation materials, the system writes meeting agendas, prepares staff reports, helps employees anticipate elected officials' questions, and drafts policy language.

The service starts at $25,000 annually and is currently deployed in municipalities including Culver City, California, and Aspen, Colorado. The company also promotes the potential for "national-level analysis" by aggregating data across neighboring jurisdictions.

Current status and security concerns

San Diego has not signed a contract or initiated pilot testing. City spokesperson Anthony Santacroce told Axios San Diego the municipality is "evaluating solutions from Madison AI that offer potential efficiencies for public records requests and legislative document information."

Public records indicate serious interest. IT assistant deputy director Serena Blanks wrote to Madison AI in late 2025 expressing desire to advance the contract, with data security identified as the remaining obstacle. Madison AI subsequently provided security documentation and completed a technology questionnaire.

UCSD communication and data science professor Stuart Geiger identified risks in uploading public documents to third-party systems. While Madison AI's security overview states customer data is never used to train or improve AI models, the company's promotional materials simultaneously advertise data-sharing capabilities across municipalities.

Democratic process implications

UCSD communication professor Lilly Irani argues the approach treats policy creation as manufacturing rather than democratic deliberation. "Democracy means we've got to take time to talk to each other," Irani told Axios San Diego, noting that policy language must withstand legal scrutiny.

Geiger warned that AI errors can compound when systems build on their own previous outputs. "If there's one little error here, one little error there, then because the AI relies on what it's already built, it can cascade, especially if people aren't really watching it," he said.

Madison AI did not respond to requests for comment.

These details were first reported by Claire Trageser for Axios San Diego.