Civic Assembly develops AI guidelines for local government
How should a local authority use artificial intelligence (AI) without losing sight of transparency, data protection and democratic oversight? In the spring of 2026, the Snohomish County Civic Assembly in the US state of Washington addressed this question. Instead of expert panels or traditional public hearings, the county opted for a process that has gained international prominence in recent years: a randomly selected Civic Assembly.
Over the course of three weekends, 29 randomly selected residents came together. The group, selected according to criteria such as age, gender, background and place of residence, worked together for around 40 hours to draw up recommendations for the future use of AI in the county administration. The findings are intended to serve as a guide for the County Council.
Viable recommendations developed collaboratively
The Civic Assembly was organised and facilitated by the National Civic League. Participants first received introductions to technical fundamentals, listened to experts with different perspectives, and then discussed the issues in facilitated small groups. The aim was not to reach a majority vote on individual positions, but to collaboratively develop viable recommendations.
The Snohomish County Council had already agreed, before the assembly began, to officially receive the proposals from the randomly selected assembly and to respond to them. In its statement, the County Council emphasised that conventional participation processes often attract primarily particularly active or vocal interest groups, whilst a randomly selected Civic Assembly could represent the population much more effectively.
Media reports as the catalyst
The assembly was prompted by media reports on the use of ChatGPT in local government, which had raised questions about transparency and democratic oversight. “It was precisely this coverage that provided the impetus for this Civic Assembly on this particular topic,” says Jillian Youngblood, director of the National Civic League’s “Civic Genius” programme.
“We mailed out 15,000 letters, combining lists from the voter file and the tax records,” recalls Jillian Youngblood, executive director of Civic Genius. “And then we hand delivered about 250 to social service organizations and to direct care providers, so we make sure that we're hitting on the full population including, maybe, people who didn’t have permanent addresses.”
Visit by former Prime Minister
Among the experts who spoke at the Civic Assembly were technology experts from the University of Washington specialising in artificial intelligence, a law professor from the University of Seattle, a city councillor who also works for Dell Technologies, and other figures, including Al Hleileh, co-founder and CEO of Civic Marketplace and an AI expert.
They were joined by the former Norwegian Prime Minister Erna Solberg, “who happened to be in town and actually has a lot to say about the regulation of technology and AI in Europe and Norway; so she came along, gave a talk and answered questions in a Q&A session about how regulation works elsewhere,” recalls Youngblood.
“Not the questions I usually hear”
AI expert Hleileh was deeply impressed by the way the audience he was speaking to approached the topic.
“The group's questions were not the questions I usually hear in government technology circles,’ he says. “They were not about procurement timelines, vendor selection criteria, or interoperability standards. They were more fundamental - and more searching - than that.
“They wanted to know whether AI systems could be wrong specifically, not just wrong in the aggregate. They wanted to know what happens to their data when a vendor is sold or acquired. They wanted to understand who is accountable when an algorithm influences a decision about a permit, a benefit, or a risk score - and whether they would ever know that an algorithm was involved at all.
Clarity among assembly members
The residents in that room understood, instinctively, that the question was not whether AI is capable of improving public services - they could see that it was. The question was who it serves, under what conditions, and what is the recourse when it gets things wrong.
“What struck me most was a specific kind of clarity that civic assembly participants tend to develop. Because they have no institutional position to defend, no budget to protect, and no procurement relationship to maintain, they are free to ask the questions that practitioners sometimes avoid.”
Enabling innovation – but under clear rules
The Civic Assembly did not, in principle, oppose the use of artificial intelligence in the district administration. Instead, it recommends a responsible, phased approach with binding guidelines.
Key recommendations include:
Human responsibility remains crucial: Decisions affecting citizens’ rights or benefits should not be made exclusively by AI systems. People should bear responsibility at all times.
High transparency: The administration should disclose when AI is being used, which systems are employed and what tasks they perform. Citizens should be able to understand when AI is involved in administrative processes.
Data protection and data security: Particularly sensitive personal data should only be processed under strict conditions. The mini-public recommends clear data protection standards as well as regular reviews.
Independent oversight: AI systems should be regularly reviewed for errors, bias, fairness and security. The recommendations provide for ongoing checks and evaluations.
Gradual introduction: New applications should first be tested in pilot projects and only rolled out more widely following a successful evaluation.
Citizen participation: Citizens should continue to be involved in fundamental decisions regarding the use of AI. The Civic Assembly recommends regular reports and ongoing dialogue formats.
Staff training: Employees of the district administration should be trained in the use of AI systems so that they can competently assess the opportunities and risks.
Overall, the recommendations reflect a cautious approach to innovation: AI should support administrative processes but must not replace democratic accountability. On 8 July 2026, the Civic Assembly’s final report was presented to the Snohomish County Council.
Close collaboration
Several members of the randomly selected assembly described the deliberations as an extraordinary experience.
Participant Jessica Rhodes emphasised the ethical dimension of the debate: „Humanity needs to be put before technology, and that's just not happening.“
Theo Moriarty highlighted the practical benefits of the work: „It's these three weekends that are intense, but my hope was to give something great to the county council government.“
“Being part of something historic”
For Hillary Moralez, the democratic significance of the project was the main focus: „To be part of something that is so new to Snohomish County, and be part of something kind of historic.“
These statements illustrate a frequently described effect of mini-publics: through intensive information and facilitated discussion, people with no prior political experience develop greater trust in collective decision-making processes.
A model for future public participation?
Ultimately, it is the county council that will decide whether the assembly’s recommendations are implemented. Regardless of the political outcome, however, the project demonstrates that complex issues for the future, such as the use of artificial intelligence, need not be left exclusively to experts or political decision-makers. Citizens selected at random – following in-depth information and structured deliberation – are able to develop nuanced and practical proposals.
The Civic Assembly in Snohomish County thus provides an example of how citizen participation through random selection can also be used in relation to future technological issues: not as a substitute for parliamentary decisions, but as a well-founded complement to the political decision-making process.
Note: This article was written with the help of AI.