Notes

Notes on AI, privacy and public records.

Short, sourced notes on AI, privacy and public records in local government, written by someone who keeps the records, and built on primary public documents.

Follow by RSS

Research notes

All 10 notes.

Newest first, each built on primary public documents.

  1. Note 10

    Every Bergen County town has until April 26, 2028

    The federal web accessibility rule gives every Bergen County municipality until April 26, 2028, to meet WCAG 2.1 AA, posted documents included.

    Topics: Web accessibility · Public records · Meetings and minutes

    Read in 3 minutes

    5 sources

  2. Note 9

    Your staff is already using AI

    Surveys this year found more local offices using AI than governing it. The first step is not a ban but a register of what’s already in use.

    Topics: AI policy · Public records · Privacy

    Read in 2 minutes

    5 sources

  3. Note 8

    Buying AI under New Jersey’s purchasing rules

    When quotations, bidding, business registration and pay-to-play apply to an AI purchase in New Jersey, what the federal OneGov offer means and how to set up a pilot a purchasing agent can sign.

    Topics: Procurement

    Read in 3 minutes

    7 sources

  4. Note 7

    Is an AI chat an OPRA record?

    New Jersey’s records law and AI guidance give offices a reason to keep chats they can find, while disclosure depends on the record and any exemption.

    Topics: Public records · Privacy

    Read in 2 minutes

    4 sources

  5. Note 6

    AI redaction and Daniel’s Law

    A redaction assistant can speed up a records response. It can’t tell whose address Daniel’s Law protects, and a covered text box isn’t a redaction.

    Topics: Privacy · Public records · Human review

    Read in 2 minutes

    2 sources

  6. Note 5

    Three things “we use AI” can mean

    An experiment, an approved pilot and an operational service are different things, with different stakes for residents, staff and public records. A claim should say which one it means.

    Topics: Status and claims · Public records

    Read in 2 minutes

    1 source

  7. Note 4

    A correct source matters more than a fluent answer

    An answer from the wrong version of a record can sound right and still be wrong. A worked example shows what to check before trusting one.

    Topics: Records and search · Evaluation

    Read in 2 minutes

    1 source

  8. Note 3

    Where a person has to stay in the minutes workflow

    Minutes are the official record of what a public body did. A tool can help prepare them, but a person has to decide what happened.

    Topics: Meetings and minutes · Public records · Human review

    Read in 2 minutes

    2 sources

  9. Note 2

    A public record isn’t automatically safe to hand to a tool

    Some personal details in public records, such as certain home addresses, are protected by law. Before a record goes into an AI tool, someone should check what’s in it.

    Topics: Privacy · Public records

    Read in 2 minutes

    2 sources

  10. Note 1

    Five questions to ask before an AI product demo

    Five questions for a vendor before any demo: what happens to the data, what is kept, who can see it, what can change and how to leave.

    Topics: Procurement · Privacy · Public records

    Read in 2 minutes

    4 sources