AI, Automated Systems, and Hudson: Briefing Document

Prepared as the shared factual baseline for a community deliberation process. This document does not take a position on any of the questions it describes — it exists so that everyone starts from the same facts before forming one. Where a claim comes from a specific person or source rather than being independently verified, that's noted.

Status: draft, pending balance review. A briefing document of this kind is only genuinely neutral once it's been checked by people who actually disagree with each other — not just written carefully by one person. This draft was prepared from public reporting and sourced facts, and includes a section presenting the strongest good-faith case on each side of the questions where Hudson residents are likely to genuinely disagree. It has not yet been reviewed by an advisory group representing contrasting viewpoints on this issue (for example, a public-safety/law-enforcement perspective and a privacy/civil-liberties perspective, at minimum). That review, and that group's agreement that this document is fair, is a required step before this should be treated as a finished briefing — not a formality to skip.

The bigger pattern

Over the past few years, automated systems have been showing up in the ordinary products and services people use — often without a clear announcement, a real choice to opt out, or even reliable notice that it happened. A few recognizable versions of this:

Decisions that used to involve a person now involve an algorithm. Hiring, lending, insurance pricing, and policing all increasingly lean on automated tools to flag, score, or recommend — sometimes with a person still reviewing the output, sometimes not, and usually with little visibility into how the tool actually reached its conclusion.
AI features get built into everyday products by default. Customer service chatbots, recommendation systems, and data collection used to train other systems are frequently switched on as the default, not something a customer affirmatively signs up for.
Private infrastructure starts doing public work. A tool bought by a private company, a landlord, or an individual resident can end up performing a function that looks a lot like policing, hiring screening, or public monitoring — without ever going through the public process that would normally govern a city or county doing the same thing.

None of this is unique to any one technology or any one town. It's the backdrop this series exists to talk about: as AI and automated systems keep arriving in more of daily life, largely decided elsewhere and discovered later, what say does a community actually get?

Hudson's first concrete encounter with this pattern is a license plate reader — camera technology that puts several of the dynamics above into one visible, local example. That's the case study below.

Hudson's case study: License Plate Readers

What an ALPR actually does

An automated license plate reader (ALPR) is a camera system that photographs every vehicle that passes it — plate, make, model, and color — and logs the time and location, regardless of whether there's any reason to suspect the driver of anything. It's not aimed at wanted vehicles; it records everyone.

The key difference from a doorbell camera: a doorbell camera's footage sits passively until a person decides to look at it. An ALPR feeds directly into a searchable database. The system is built to be queried — by time, by location, by plate — and by design, the public has no visibility into who is running those searches or why.

Flock Safety is one manufacturer of ALPR systems; it markets itself on public safety grounds — recovering stolen vehicles, finding missing persons, and supporting Amber/Silver Alert response. That pitch has gotten Flock cameras into more than 6,000 communities nationwide. Axon is a different manufacturer that also sells ALPR technology used by the Hudson Police Department (more on the distinction below).

Flock sells add-on products that expand what its system can do:

Nova cross-references license plate data against other law enforcement databases and commercial data brokers (companies that buy and sell personal data — where you live, who you're related to, what else is known about you online) — so a search can go from "a car passed by" to "who owns that car and who they're connected to."
FreeForm lets a user search with plain-language description ("a red pickup truck with a dog") — meaning the system's AI is analyzing footage of people, not just plates.

Data sharing: what "National Lookup" means

Flock's default local retention is 30 days, after which footage is automatically deleted — the company describes this as a privacy safeguard.

But that number only applies to data that stays local. Roughly three out of four agencies using Flock have a feature called National Lookup switched on. When it's on, a local camera becomes a node that any other participating agency in the country can search — and a search from elsewhere can copy that data out before the local 30-day window closes.

Reporters have documented license plate data captured by Flock cameras reaching Immigration and Customs Enforcement, sometimes through formal data-sharing agreements and sometimes through local police running searches on ICE's or Border Patrol's behalf without residents' knowledge. Flock's stated position is that it no longer contracts with ICE, dropped federal agencies from National Lookup, and blocks immigration-related search terms — changes the company made after reporters exposed the earlier practice.

Documented misuse isn't limited to immigration enforcement: a Texas sheriff ran a nationwide Flock search looking for a woman who'd had an abortion; New York agencies have been found running plate searches tied to protest activity.

Accuracy and security

False positives: Los Angeles data suggests roughly one-third of the system's stolen-vehicle plate matches were false positives — meaning innocent drivers were pulled over and treated as felony suspects because a camera misread a plate.
Unsecured cameras: In December, researchers found more than 60 Flock cameras livestreaming to the open internet with no login required, discoverable through a Google search — including at least one documented case where an outside party could pan, tilt, and zoom the camera itself.

What's known about Hudson and Columbia County specifically

Facts below are drawn from public reporting ("Flock Cameras in Hudson and Beyond," All My Dead & Living Things, Aug. 21, 2026) and named officials' public statements. Some open questions genuinely have no public answer yet — those are marked.

Hudson's only camera reported to DeFlock (a volunteer-run public mapping database) is privately owned, mounted at 441 Warren Street, on a property tied to Hugo van Vuuren. There is no written record confirming whether this camera has ever shared data to any outside agency — this is an open question, and would require a Freedom of Information Law (FOIL) request to resolve.
The City of Hudson has no contract with Flock. Mayor Joe Ferris confirmed this directly. The Hudson Police Department does use license plate readers made by Axon, a different manufacturer. Ferris has stated that "Axon has significantly stronger safeguards including no national data-sharing networks" than Flock — this is the Mayor's characterization, not an independently audited comparison, and worth reading as such rather than as settled fact.
Columbia County's ALPR history is more contested. The previous sheriff, Don Krapf, purchased Flock software and cameras in 2025 — after the Board of Supervisors' Public Safety Committee had already tabled his request for a formal agreement with Flock back in March 2023. He used part of a $279,790 state law enforcement grant to buy the software in December 2025 and order the cameras that October. The county attorney's office is currently investigating whether that purchase was valid without full Board approval — this is an open, unresolved question, not a settled one.
Current Sheriff Jackie Salvatore has referenced two additional privately-owned Flock cameras in the county (one not yet reported to DeFlock), plus ten cameras belonging to the Columbia County Sheriff's Department — including one at Greenport Town Park (a central gathering spot for kids countywide), several in the Lowe's parking lot in Greenport, and one near Cantele's Laserwash on Fairview Avenue, facing Hudson's 5th Ward.
DeFlock shows 15 reported cameras countywide as of this reporting — fewer than neighboring Dutchess (125) or Albany (122) counties. Because DeFlock relies on volunteer reporting, this may undercount what's actually installed.
Mayor Ferris has said he supports pending New York State legislation, A.10808/S.9890 (the "Comprehensive ALPR Guardrails" bill), sponsored by Senator Michelle Hinchey and Assembly Member Pamela Hunter, and is looking at whether similar guardrails could be implemented at the city level while the state bill is pending.
There is currently no Hudson municipal policy governing ALPR use.

How nearby communities have handled this

Troy, NY: Spent a year working toward public rules on its 26 Flock cameras. A proposed local law would have restricted use to outstanding-felony-warrant cases; it drew pushback from the county DA and police chief and didn't pass as written. The city landed on a negotiated 60-day working period between the mayor, council, DA's office, and police, resulting in policy changes (including turning off the national search tool) rather than a binding law.
Saranac Lake, NY: Installed a dozen Flock cameras via state grant in 2024. After residents learned of it, the village board halted the program, shut off the cameras, and voted 4–1 to ban them outright.
Trumansburg, NY: Simply chose not to renew its Flock contract when it came up for renewal.

Where residents are likely to genuinely disagree

The facts above are not in serious dispute. What follows are the two open value questions this briefing exists to prepare you for — presented as the strongest good-faith case on each side, not as a recommendation. Reasonable people, with the same facts, land differently on both.

Should Hudson have any license-plate-reader system at all — city or county-level?

The case for

ALPRs can help recover stolen vehicles, find missing persons, and speed response to Amber and Silver alerts — the reason more than 6,000 communities have adopted some version of this technology. Hudson residents' plates are also very likely already being read by camera networks just outside city limits (Columbia County's own cameras, including ones at Greenport Town Park and near Cantele's Laserwash facing the 5th Ward) — declining to have any Hudson-controlled system doesn't necessarily mean Hudson residents avoid this kind of data collection, it may just mean the data is collected by an agency Hudson has no say over.

The case against

The technology collects data on every driver regardless of suspicion, not just people under investigation — a fundamentally different posture than traditional policing. Nationally, this data has documented histories of reaching ICE, being used in a nationwide search related to an abortion, and being used to track protest activity — none of which required a local department to intend that outcome. Roughly a third of stolen-vehicle plate matches in Los Angeles data were false positives, meaning innocent drivers were treated as felony suspects. And once a system is installed, added capability (like Flock's Nova and FreeForm products) can expand what it does well beyond the original stated purpose, without a new public conversation. Saranac Lake installed Flock cameras and later voted 4–1 to ban them once residents examined what the system actually did.

Is Axon meaningfully safer than Flock, given Hudson Police already use it?

The case for

Mayor Ferris has stated Axon has no national data-sharing network — and the most serious nationally documented harms (data reaching ICE, the abortion-related search, protest-related tracking) all trace specifically to Flock's National Lookup feature. A system whose data stays local, under an accountable local department, is a structurally different risk than one that can be searched by any participating agency in the country.

The case against

That claim about Axon's architecture is the Mayor's characterization; no independent audit or public documentation has confirmed it the way Flock's data-sharing practices have been reported on by journalists. Axon is still an ALPR — it still logs every vehicle's location regardless of suspicion, and nothing publicly confirms it's exempt from the same accuracy concerns (false positives) documented for the technology broadly. And without any Hudson municipal policy governing retention, audit, or permitted use — for Axon or any future vendor — the actual safety of the system rests on the vendor's word and internal department practice, not on an enforceable local rule that would survive a change in vendor, department leadership, or company policy.

A short glossary

ALPR: Automated License Plate Reader; the general category of technology (Flock and Axon both make ALPR products)
National Lookup: Flock's feature that lets a participating agency search camera data from other participating agencies nationwide
MOU: Memorandum of Understanding; a formal (though not always legally binding) agreement between two parties, such as a police department and a vendor
FOIL: Freedom of Information Law; the New York State process for requesting public records
Data broker: A company that collects and sells personal information, often without the person's direct knowledge
DeFlock: An independent, volunteer-maintained public map of reported ALPR camera locations nationwide; not an official government or company database, and therefore likely incomplete

Sources: "Flock Cameras in Hudson and Beyond," All My Dead & Living Things (Substack), Aug. 21, 2026; public statements by Mayor Joe Ferris; Columbia County Board of Supervisors Public Safety Committee records.

← Back to Can We Talk About AI?