AI use policy template for small cities and towns (free, no IT director required)

Your staff are already using AI. Surveys of local-government officials consistently find ChatGPT and similar tools in daily use for drafting letters, summarizing documents, and answering resident questions — usually with no rules in place. Roughly two-thirds of cities report they have adopted or are developing an AI policy, but those are mostly larger governments with IT departments and attorneys on staff. As Route Fifty put it, America's “16,000 small towns are on their own.”

This page is our answer: a full AI use policy template you can read, copy, and adapt today. It assumes no IT director, no procurement office, and no budget — just a clerk or administrator who needs sensible rules before something goes wrong. The riskiest posture is not “AI with rules” or even “no AI” — it is what most small towns have now: quiet, unguided AI use with public records and resident data in play.

The template: AI Use Policy for [City/Town of ____________]

Bracketed text is yours to fill in. “Policy Administrator” means your town clerk or administrator unless council names someone else.

1. Purpose and scope

This policy governs the use of artificial intelligence (AI) tools — including generative AI services such as ChatGPT, Microsoft Copilot, Google Gemini, and Claude — by employees, elected and appointed officials, contractors, and volunteers of the [City/Town of ____________] (“the Town”) in the course of Town business.

It applies to AI use on Town-owned devices and accounts, and to AI use on personal devices or accounts whenever the resulting work product is used for Town business. This policy is written for a town that may have no IT department: wherever it assigns a duty to the “Policy Administrator,” that duty belongs to the [Town Clerk / Town Administrator] unless the Council designates someone else.

2. Definitions (plain English)

• Artificial intelligence (AI): software that produces text, images, summaries, predictions, or recommendations by finding patterns in data, rather than by following fixed rules a person wrote.
• Generative AI: AI that creates new content — drafts, summaries, images, translations, computer code — in response to a prompt. ChatGPT, Copilot, Gemini, and Claude are generative AI.
• Public AI tool: any AI service operated outside the Town's control where prompts and uploads may be stored, reviewed, or used for model training by the vendor. Free consumer accounts — including free ChatGPT accounts — are public AI tools.
• Approved AI tool: a tool listed on the Town's Approved AI Tools List (Section 9) and used within the limits recorded on that list.
• Non-public information: any information the Town holds that is not already freely available to the general public — including anything that would be redacted from or withheld in response to a public-records request.

3. Permitted uses

Subject to human review (Section 6) and the prohibitions in Section 4, staff may use approved AI tools to:

1. Draft and edit routine documents: letters, public notices, newsletters, job postings, website copy, and social media posts.
2. Summarize documents that are already public: ordinances, agendas, minutes, budgets, and published reports.
3. Conduct research and brainstorming, provided every fact, figure, citation, and legal reference is verified against an authoritative source before it is used.
4. Reformat, proofread, or restructure text and data that contain no non-public information.
5. Draft internal procedures, checklists, and training materials for staff review.

4. Prohibited uses

Staff may never:

1. Enter non-public information into a public AI tool. This includes — without exception — Social Security numbers and other personally identifiable information, personnel and medical records, law-enforcement and court records, juvenile records, utility-customer account data, and any record or portion of a record that is not public. Never paste resident PII into a public AI tool.
2. Use AI as the sole basis for any decision that affects a resident's or employee's rights, benefits, property, or employment — including permits, citations, code enforcement, hiring, discipline, and benefits eligibility. A named human being makes the decision; AI may only assist.
3. Use AI to impersonate any real person or organization, or present AI output as the personal work of someone who has not reviewed and adopted it.
4. Use AI-generated content as legal advice or in place of review by the Town attorney.
5. Sign up for, purchase, or connect AI tools to Town systems outside the procurement process in Section 8.

5. Public records and retention

Treat AI prompts and outputs used in Town business as public records under [state] public-records law. A record that would be public if typed in a word processor does not become private because AI helped produce it. Staff must:

• Save AI-assisted work products in the Town's normal records locations (shared drives, the records system, minute books) — never solely inside an AI chat history.
• Apply the Town's existing retention schedules to AI-assisted records, based on the content of the record, not the tool that produced it.
• Remember that chat histories in consumer AI tools are not a records-management system and may be deleted or changed by the vendor at any time.

6. Accuracy and human review

Generative AI produces confident-sounding errors (“hallucinations”), including invented facts, numbers, quotations, and legal citations. Every AI-assisted work product must have a named human owner who reviews it and is responsible for its accuracy before it is used, published, sent, or relied on. The owner verifies names, dates, figures, and every citation. “The AI wrote it” is never an acceptable explanation for an error.

7. Transparency and attribution

• Disclose AI assistance when a published document or communication issued in the Town's name is substantially AI-generated.
• Always disclose when residents are interacting directly with an AI system (for example, a website chatbot).
• Staff need not label routine internal drafting assistance, but must answer honestly if a resident, official, or court asks whether AI was used.

8. Procurement and vendor AI

Before the Town buys — or accepts a free trial of — software with AI features, the Policy Administrator obtains written answers from the vendor to these questions:

1. What Town or resident data does the AI feature use, and is that data used to train models that serve other customers?
2. Where is the data stored, and who at the vendor can access it?
3. Can the AI features be disabled without losing the rest of the product?
4. What happens to the Town's data when the contract ends?
5. How does the vendor detect, correct, and notify customers about AI errors?

No resident or employee data may be shared with a vendor's AI system without a written contract covering confidentiality, permitted data use, and deletion.

9. Training and the Approved AI Tools List

The Policy Administrator maintains a one-page Approved AI Tools List recording, for each tool: the tool and plan name, approved uses, prohibited uses, and the date approved. Start with one or two tools — preferably ones included in software the Town already pays for. A starter table:

All staff receive brief training on this policy — even thirty minutes at a staff meeting — before using AI tools for Town work, and new hires receive it within [30] days of starting.

10. Incident reporting

If non-public information is entered into a public AI tool, or AI-generated content containing a significant error is published or relied on, the employee involved reports it to the Policy Administrator the same business day. Good-faith self-reporting is not punished; the goal is to fix problems quickly. The Policy Administrator documents the incident, consults the Town attorney where legal obligations may apply, and updates training or the Approved AI Tools List as needed.

11. Review cadence

The Policy Administrator reviews this policy at least annually — and after any reported incident or significant change in state law or guidance — and reports to the Council on AI use and incidents at least once a year.

12. Model adoption resolution (fill in the blanks)

This is the part clerks actually need: a short resolution your council can pass to adopt the policy. Paste it into your normal resolution format.

Download .docxDownload as Markdown (.md)

State notes: South Carolina and North Carolina

How this template was built

Reviewed against published municipal AI policies (San José, Seattle, Boise, Bellevue) and association guidance, then rewritten in plain English for towns without an IT department. Adapt it to your state's law. It is not legal advice — have your city attorney review the policy before your council adopts it.

Frequently asked questions