As AI creeps into Napa County government, leaders hit pause on policy — for now – The Press Democrat
Artifical intelligence is already at work in Napa County, but supervisors say the public deserves to know more before setting the rules.
Artificial intelligence is already quietly helping Napa County officials translate public meetings, design materials and streamline daily work, but there’s still no official policy on how this powerful technology can be used or what limits should be in place.
This week, county supervisors hit pause on a proposed set of rules for generative AI — technology that creates new text, images, audio or other content based on user prompts — saying the draft policy didn’t go far enough to explain how the tools are already being used behind the scenes.
Supervisor Anne Cottrell supported creating safeguards around AI use but said she also wanted the policy to help educate the public, not just govern internal operations. Tuesday marked the board’s first public conversation about AI, she noted, and more discussion is needed.
Greg Bown, the county’s interim chief technology officer, said the policy is meant to help avoid the common pitfalls tied to AI. The county is particularly concerned about data leakage, copyright violations, and the spread of misinformation or bias. Bown said a major pitfall of AI is over-reliance on it, and trusting the output of the systems without adequate human review.
“Society has developed an endemic automation bias by which humans favor suggestions from automated decision-making systems blindly, often ignoring their own better judgment,” the policy states. “Using output from GenAI tools without reviewing it for accuracy places the County at risk and may harm Napa County’s reputation with the public and employees.”
Other counties, including Sonoma, Santa Cruz and San Francisco, have already adopted similar AI policies in recent years.
Bown added that a major pitfall of AI is over-reliance — trusting the output of these systems without adequate human review. Governments that have deployed automated or algorithmic tools have repeatedly run into trouble.
In one example, a New York City chatbot powered by generative AI gave small business owners incorrect — and in some cases illegal — advice, including guidance that violated housing and labor laws, according to an investigation by The Markup.
In another case, the California Employment Development Department used AI tools to flag potentially fraudulent unemployment claims. In late 2020, it paused benefits for 1.1 million people, but a later state analysis found that more than half of those claims, around 600,000, were legitimate.
Even as Napa County’s policy remains under revision, the board will now allow county departments to begin using some AI tools that have been on hold, as long as safeguards are in place and a new digital governance committee oversees decisions.
That move raised questions for some supervisors, including Belia Ramos, who warned that parts of the government may already be using AI in ways that could fall out of bounds once the policy is finalized.
“What I don’t want to do is adopt a policy where something that we are currently utilizing is not in compliance with our policy,” Ramos said.
The five-page draft policy, a year in the making, lays out what generative AI is, what county workers can and can’t do with it, and how to avoid exposing sensitive information to tools like ChatGPT. It also requires human oversight for any AI-generated content.
The draft was shaped by input from local departments and national organizations, including the National Association of Counties and the Government AI Coalition, a San Jose-led initiative focused on responsible public-sector AI use.
But supervisors said the document felt too dry and too technical, and didn’t address deeper concerns about how AI could change government operations, decision-making or public communication in the long run.
Supervisor Amber Manfree said she appreciated that the policy was framed as a starting point, but emphasized that human oversight will be essential to maintaining the quality of public service.
Assistant County CEO Becky Craig said departments would be “careful and selective” in moving forward, and that nothing would be approved without strict vetting.
The county’s digital innovation officer Meg Ragan noted AI is already baked into tools like Canva and Wordly, and said the county is following national conversations on government AI use.
“Developing a responsible AI program will have many layers, and this policy lays a foundation to manage risk, protect sensitive data and uphold the security and integrity our community expects,” she said.
You can reach Staff Writer Edward Booth at 707-521-5281 or edward.booth@pressdemocrat.com.
