It’s easy to assume that if a document is public, any tool can use it. In municipal work, that assumption doesn’t hold for long.
A land use application is a public record, and it can include an applicant’s phone number and signature. Meeting minutes can name a resident who spoke.
Daniel’s Law
New Jersey’s Daniel’s Law protects “the home address or unpublished home telephone number” of covered persons. The State’s Office of Information Privacy works with state, county and local agencies to shield that information on their own websites, and each agency appoints or signs up redactors to do it.1
That changes what “public” means in practice. An address that was properly redacted on a website can still be sitting in the original file. If that file goes into an AI tool, the tool can’t tell which details are protected, and it can repeat any of them.
Three questions before a record goes into a tool
- What’s actually in it? Look past the type of document to its contents: names, addresses, signatures, phone numbers and personal circumstances.
- Who will see the output? A tool one staff member uses is a different risk from one that answers the public.
- Where does it go, and how long does it stay? Ask what the service keeps, where it keeps it, who can see it and whether that copy becomes a record the agency has to manage.2
Joining records creates something new
One document shows an address. Another shows when someone is away. A third has a phone number. Each is public. A tool that joins them into one profile has made something no single record contains. That calls for care, even when every source is technically public.
The rule I follow
I work with the smallest set of public documents that answers the question, with personal details removed, or with synthetic material. If I can’t be sure what a document contains, it doesn’t go in.
Sources
-
New Jersey Department of Community Affairs, Office of Information Privacy (Daniel’s Law). Accessed September 24, 2026. ↑
-
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. Accessed September 24, 2026. ↑