Shawn Landres PhD, a Senior Fellow at the UCLA Luskin School of Public Affairs, is a second-term Santa Monica Planning Commissioner and former chair of the Planning and Social Services commissions. The views expressed here are his own.
Tonight the City Council takes up a proposal from Mayor Torosis and Councilmembers Hall and Zwick to make Santa Monica a national model for responsible AI use in local government. The item would update the City’s AI policy, authorize enterprise AI accounts, embed AI training in the reestablished Santa Monica Institute, and commit the City to protecting workers from displacement.
Clearly there is a pressing need to address how the City uses AI. But AI already is reaching the City: through the public comments, letters, and increasingly the phone lines that shape every decision the Council and its commissions make.
Council ought to consider not only how the City uses AI, but also how AI might be using the City.
Last December, the Planning Commission unanimously took up my call to delay a vote on housing subdivision rules because the public had not had enough time to weigh in. I told residents: “Tell us what you think. We will read your comments and take them seriously, all of them, even if we disagree with them.”
They did. In advance of our March 4 study session, we received more than 250 public comments, many within 24 hours of the meeting. I analyzed the packet with AI, asking which letters departed from organized templates and which adapted the template to the writer’s own concerns. AI helped me do my job better. The same technology that helped me read 250 comments also could have written them.
In February, the LA Times revealed that a consultant used CiviClick, an AI platform, to generate over 20,000 opposition messages to the South Coast Air Quality Management District, targeting clean air rules expected to prevent 2,500 premature deaths. Three of five purported commenters had no knowledge of the letters. The board voted 7-5 to reject the rules. When residents cannot trust that public comments come from actual people, civic engagement stops connecting communities to decision-makers and starts connecting algorithms to the public record.
The quieter problem may prove harder to fix. A Stanford-led study found that every large language model tested for summarizing public comments scored majority positions higher than minority ones. It is one thing for a disabilities commission to use AI to disaggregate sidewalk mobility experiences — that is precision. It is bias when an AI tool, in search of coherence at the expense of accuracy, treats more numerous homeowner comments as more representative than renter comments, or flattens the comments of Spanish-speaking families and workers into a summary that mirrors the majority view.
Five reforms would address how AI reaches City Hall from the outside. First, require any body using AI to summarize public comments to disclose the tools and parameters used to build the summaries; the Little Hoover Commission’s 2024 report calls for this transparency. Second, publish a model anti-bias prompt and agendize it for discussion and approval; the choices in a summarization prompt are policy decisions, and policy decisions belong in open session. Third, mandate disclosure of AI-generated comments, following campaign finance law’s source-tracing approach for out-of-state contributions. Fourth, track AI equity: who participates, through which channels, and whether the process is reaching the residents it was designed to serve. Fifth, design remote public comment to be easy for humans and hard for AI agents. AI-generated callers eventually could attempt live public comment, but California law prohibits requiring commenters to identify themselves. Friction-based measures like rate limiting and live-connection requirements could help, but the state will need to address this tension directly.
Santa Monica can lead here. The Council’s Tuesday item establishes how the City uses AI responsibly within its own walls. A companion set of policies could establish how the City handles AI when it arrives from outside: in the comment packet, on the phone line, in the summary of what our neighbors said. SB 707 compliance itself belongs in that companion framework: after July 1, AI very likely could be translating agendas, captioning meetings, and processing comments on behalf of the City, and residents deserve to know which tools are doing that work and under what rules. Together, the two frameworks would do what neither does alone: protect the integrity of the conversation between residents and the government that serves them.
Author
About The Author
