For most offices, the real question is whether anyone has written down how AI is already being used.
Use runs ahead of rules
In a Tyler Technologies survey reported by ICMA in May 2026, 33 percent of 109 public-sector leaders, 54 of them municipal officials, said they had a defined AI policy or governance framework in place. Yet 44 percent were already experimenting with generative AI for uses such as chatbots and content drafting. The municipal respondents named their biggest barriers: limited internal expertise or staffing (63 percent) and uncertainty about the ethical or legal implications (52 percent).1
Granicus’s 2026 survey of government organizations found the same pattern: 55.7 percent using AI and 42.9 percent with formal AI policies in place. Among local agencies, 57.1 percent reported using it.2
Why a ban doesn’t close the gap
Generative AI now arrives inside software that offices already pay for. As the National League of Cities puts it, that can make “the line between approved and unapproved use unclear.”3 A ban on AI tools doesn’t reach a feature that switched on in last month’s update. And it pushes the rest into personal accounts, where the office can’t see the work or keep its records.
That matters because New Jersey’s records guidance is plain: “Records associated with or created by governmental systems using AI/ML are therefore subject to the State’s public records law.”4 An office can’t manage records it doesn’t know exist.
Start with a register
The National League of Cities’ advice fits in eight words: “Know where and how AI is being used.”3 A register does exactly that, with one row per tool:
- the tool and the department using it;
- what it’s used for;
- the data it can see;
- who approves what it produces;
- the records it creates and how long they’re kept;
- the date someone last reviewed the row.
Then a policy has something to govern. New Jersey’s interim guidance for State employees is a good model for the rules that follow: label content made with generative AI, keep personal information about residents, colleagues and yourself out of the tools, and verify everything before it’s used.5
The AI use register in my toolkit has the columns listed above, ready to print or open as a spreadsheet. What goes into your policy is your governing body’s decision, on your attorney’s advice; the register gives it the facts.
Sources
-
Burgandi Grace, Tyler Technologies, “AI at the Local Level: What Municipal Leaders Told Us,” ICMA, May 4, 2026. Accessed September 27, 2026. ↑
-
Dan Firrincili, “From policy to practice: How AI is quietly reshaping government operations in 2026,” Granicus, April 6, 2026, reporting Granicus’s 2026 State of Digital Government survey. Accessed September 27, 2026. ↑
-
Connor Heaton, CivAI, “AI in Local Government: Lessons from 2025 and Emerging Cyber Threats in 2026,” National League of Cities, March 27, 2026. Accessed September 27, 2026. ↑ ↑2
-
New Jersey Division of Revenue and Enterprise Services, Guidelines for Developing Retention and Disposition Policies for Artificial Intelligence/Machine Learning Systems, February 2025, revised July 2025, page 3. Accessed September 27, 2026. ↑
-
New Jersey Office of Information Technology, Joint Circular 23-OIT-007, State of New Jersey Interim Guidance on Responsible Use of Generative AI. Accessed September 27, 2026. ↑