Missoula County adopts policy requiring human review, disclosure for AI-produced work

Missoula County administrators voted to adopt a new countywide policy governing the use of artificial intelligence in county work, requiring a human to take responsibility for AI-produced documents and disclosing when AI was used.

The policy, presented at the administration meeting, was pitched as a risk-management measure to limit county liability from errors or harms that could arise when staff rely on generative AI tools. The presenter said the policy “echoes the rest of, state, local government, municipal government” and aims to “effectively limit liability around the use of AI for … work product in … local government.”

Why it matters: County staff increasingly use generative AI that can create text, images or video. Officials said the law and liability framework governing AI are unsettled and that internal policy is the best near-term protection while legal standards develop. The presenter described the policy’s core requirement as assigning a human reviewer to attest that they have read and edited AI-produced material.

Key provisions and examples cited during discussion included a requirement that a person take ownership of AI outputs (described in the meeting as “human in the middle”), auditing or editing AI-generated content before release, and rules limiting where county data may be sent or stored. As one participant summarized, “It’s called, human in the middle.” Another said the policy should also require explicit attribution when AI tools create a product: “attribution that something is created with AI too.”

The board moved and seconded the item and approved the policy without additional amendment; a voice vote was taken and the motion passed.

Implementation and limits: Presenters noted the policy is intended to be pragmatic and to mirror common practices in other government offices. They emphasized it is not a technical standard or a legal determination about liability; rather, it is an administrative requirement that staff document human review and limit data flows. The presenters cautioned that technical details such as image- or video-generation raise additional questions and that further policy refinement may be needed as the technology and law evolve.

What comes next: With the policy adopted, staff will implement the human-review and disclosure requirements in departmental workflows; no timeline for training or enforcement was specified during the meeting.