ALPR, Explained: From Stolen-Car Tool to Mass Surveillance

“It’s just a license plate reader.” You hear it at every county meeting. It is also, by now, badly out of date. The camera on the pole is fifty-year-old technology — but what sits behind it is not. Here is the plain-English version: what these cameras are, who runs them, and how a tool built to catch stolen cars became a standing record of where everyone goes.

What an ALPR actually is

An automated license-plate reader (ALPR) is a camera plus software. It photographs a passing vehicle, reads the plate with optical character recognition, and stamps each read with a time and a location — then stores it. Modern systems record more than the plate: the car’s make, model, color, and features. One photo of one car is harmless. A stored, searchable history of every car is the point — and that history is what the vendors sell.

How long have they existed?

About fifty years. ALPR was invented in 1976 by the United Kingdom’s Home Office Police Scientific Development Branch. Working prototypes ran at the Dartford Tunnel by 1979, and the first arrest from a stolen-car hit came in 1981. The technology stayed niche until cheaper software spread it in the 1990s, and U.S. law enforcement adopted it through the 2000s. So the camera is not new. What’s new is scale, artificial intelligence, and networking — the three things that turned a roadside gadget into mass surveillance.

Who uses them?

Two worlds, and most people only know about one.

Government. Police departments, sheriffs’ offices, state agencies, and — through data-sharing — federal agencies. The vendors include Flock Safety (founded 2017, Atlanta), Motorola/Vigilant, Genetec, Rekor, Axon, Verkada, and Leonardo/ELSAG. We traced that whole chain — the land, the vendor, the cloud — in The ALPR Machine.

Private companies — the part people miss. The repossession and insurance industry runs its own nationwide plate network, separate from police. Digital Recognition Network (DRN) and Vigilant — now owned by Motorola Solutions, which bought their parent company for $445 million in 2019 — mount cameras on tow trucks and repo trucks that scan constantly as they drive. DRN’s database holds more than 9 billion plate images and adds roughly 70 million scans a month, sold to lenders, repossession firms, insurers, private investigators, and law enforcement. Parking operators, toll systems, and marketers use the same tools. A private company can build a picture of your movements you never consented to — and sell it, including to police.

And the hosts are often businesses you shop at: a national retailer, Lowe’s, gave the N.C. SBI standing, warrantless access to its Flock cameras across three states — and never told its customers (details here).

When did AI get added?

Reading the plate itself is optical character recognition — a form of pattern recognition that’s been there from the start. The real artificial-intelligence leap came in the late 2010s, when machine vision learned to recognize a vehicle’s make, model, color, and distinguishing features — bumper stickers, roof racks, dents, mismatched panels — even a car with no plate at all. Flock calls this “Vehicle Fingerprint”. That is the moment a plate reader became a vehicle-recognition system, searchable by plain description and by image.

When did a database of movements get added?

The private sector built it first. Through the 2000s and 2010s, DRN and Vigilant accumulated billions of time-stamped, geo-located scans — a commercial location database before most people had heard of these cameras. On the government side, Flock’s cloud network (from 2017 on) made a continuously-updated location history for ordinary vehicles routine — and searchable across agencies and state lines. The stored movement trail, not the single photo, is the product.

When did it go from plates to people — and faces?

Flock’s own patent answers this. U.S. Patent 11,416,545, granted August 16, 2022 to inventors Garrett Langley (Flock’s CEO) and co-founder Matt Feury, describes pulling video from many unrelated cameras into one searchable network, recognizing a specific named person and alerting a subscriber (its own example is a subscriber’s teenage daughter), and searching by “face recognition data points.” Flock says publicly its plate readers aren’t facial recognition or “live tracking.” But the capability to track people, not just plates, is written into the company’s own patent. Read the patent yourself.

What changed, and when — the short version

“These cameras are nothing new” — the view from thirty years ago

Because the technology is genuinely old, defenders often wave the concern away: officers have run license plates for decades, the argument goes, so what’s the big deal now? You could hear exactly that at the Brunswick County commissioners’ meeting on September 21, 2026, where Vice-Chairman Marty Cooke, defending the program, told the board:

“I did sit in a Motorola presentation a couple of years back where they talked about these cameras — they can even determine whether a person has a weapon. But we’re looking at it from the standpoint of an individual driving into our jurisdiction with a warrant on their car, and it flags it. The courts consider that similar to an individual officer writing that information down and running the tag.”

That one answer contains the whole problem. In a single breath he acknowledges a vendor pitched a system sophisticated enough to flag whether a person is armed — and then defends it as if it were an officer jotting a plate on a notepad. Both can’t be true. Running one plate against a hotlist is a check. What exists now is a different category of tool: a networked, AI-driven system that photographs every car, stores each sighting with a time and a place, recognizes a vehicle by make, model, and features, and lets an officer pull up a month of anyone’s movements across multiple states in seconds — no stolen car, no suspicion, no warrant. And the claim that “the courts consider that similar to… running the tag” is exactly the argument a federal judge rejected ten days later in United States v. Kyle, calling a warrantless ALPR location search “indiscriminate mass surveillance.” A radio check in 1995 could tell you whether one plate was stolen. Today’s system keeps a standing, searchable history of everyone — and calling that “the same thing we’ve always had” is a view from thirty years ago.

How many fixed ALPR cameras are in North Carolina?

No official count exists — no agency publishes one. The best public proxy is the crowdsourced DeFlock map, which has documented on the order of 2,800-plus ALPR cameras in North Carolina, roughly 85% of them Flock, across dozens of agencies (per DeFlock). Because it depends on volunteers spotting cameras, that is a floor, not a ceiling. In the Wilmington / New Hanover area alone the same map shows at least 62. When the government won’t publish the number, the public has to keep it (see the statewide map).

And the number is set to rise, not fall. In the 2026 state budget, the North Carolina General Assembly made the SBI/NCDOT statewide plate-reader pilot permanent and opened it to any agency — a local department can now petition NCDOT to mount a camera in a state right-of-way, on roads in all 100 counties, with Flock supplying most of the equipment. So even as some towns move to cancel or ban, the state is building in the other direction, and we should expect more cameras across North Carolina, not fewer. (How North Carolina’s ALPR law works.)

How many are on law-enforcement vehicles?

Nobody publishes that either — and it is growing fast. Beyond the cameras on poles, agencies are adding car-mounted ALPR: Brunswick County, for example, signed a roughly $4 million, five-year Axon agreement to put mobile plate readers in Sheriff’s vehicles. On the private side, DRN’s scanners ride on repossession and tow trucks. Every one of those is a moving camera the fixed-camera count never captures — which means the real reach is larger than any map shows.

Do the systems integrate? (the Kyle case)

Yes — integration is the whole design. In United States v. Kyle (N.D. Okla., Oct. 1, 2026), a deputy ran one out-of-state plate through two ALPR systems from his patrol-car laptop — Flock and VehicleManager — and a single query returned a month of that car’s movements: more than fifty sightings across multiple states, shared nationwide. Flock’s network is searchable across agencies and state lines; ALPR feeds real-time crime centers (Fusus/Axon) and can even dispatch a drone to a camera alert; and the private DRN database and government systems both end up in front of police. Poles, patrol cars, and private trucks aren’t separate tools. They are one interoperable grid.

Why this matters

The camera is fifty years old. The surveillance is not. AI, the cloud, and nationwide networking took a tool for catching stolen cars and turned it into a standing record of where ordinary people go — held largely by private companies, shared across the country, and searched without a warrant or any suspicion that a crime occurred. A federal judge has now said the obvious: that is “indiscriminate mass surveillance.” And the first North Carolina town has said no — Atkinson banned the cameras outright. What one town did, any can.

Quick answers

What is an ALPR camera? An automated license-plate reader is a camera plus software that photographs passing vehicles, reads their license plates, and records the time and location of each read in a searchable database. Modern systems also log a vehicle’s make, model, and color.

When were license-plate readers invented? In 1976, by the United Kingdom’s Home Office Police Scientific Development Branch. Prototypes ran by 1979, and the first arrest from a stolen-car hit came in 1981.

Who uses ALPR cameras? Police, sheriffs, and state and (through data-sharing) federal agencies — and private companies. The repossession and insurance industry, led by Digital Recognition Network and Vigilant (now Motorola Solutions), runs its own nationwide network and sells the location data.

When did AI get added to license-plate readers? In the late 2010s, when machine vision began recognizing a vehicle’s make, model, color, and features — even cars with no plate. Flock calls this “Vehicle Fingerprint.”

Does Flock track people, not just plates? Flock says its readers are not facial recognition, but its own U.S. Patent 11,416,545 (granted 2022) describes recognizing a specific named person and searching by face across unrelated cameras.

How many ALPR cameras are in North Carolina? There is no official count. The crowdsourced DeFlock map documents on the order of 2,800-plus ALPRs in North Carolina, about 85% of them Flock — a minimum, not an official total.

Are license-plate readers mounted on police cars? Yes. Many agencies add car-mounted ALPR (Brunswick County signed a roughly $4 million Axon deal for mobile readers), and private repossession and tow trucks carry them too — none of which fixed-camera counts capture.

Do ALPR systems share data nationwide? Yes. Networks like Flock’s are searchable across agencies and state lines; in United States v. Kyle (2026) one query returned a month of a car’s movements across multiple states.

Is North Carolina getting more ALPR cameras? Likely yes. The 2026 state budget made the SBI/NCDOT statewide plate-reader pilot permanent and open to any agency across all 100 counties.

Is ALPR surveillance legal? It is contested. In October 2026 a federal judge in United States v. Kyle ruled a warrantless ALPR location search unconstitutional, and Atkinson became the first North Carolina town to ban the cameras outright.

Sources

• History: Traffic Technology Today; Wikipedia — ANPR.
• Commercial/private networks: Engadget and Carscoops on DRN/Vigilant; Motorola Solutions’ 2019 acquisition of VaaS (DRN + Vigilant).
• Flock capabilities: U.S. Patent 11,416,545 B1 (Google Patents); Flock Safety company materials.
• NC counts: the crowdsourced DeFlock map (deflock.me) — a public minimum, not an official total.
• Integration & the ruling: United States v. Kyle, N.D. Okla., Oct. 1, 2026.

General information for public discussion, not legal advice. Dates and figures are drawn from the sources linked above; company capabilities are described from Flock Safety’s own patent and published materials. North Carolina camera counts come from the crowdsourced DeFlock map and are a verified minimum, not an official total — counts change. Corrections: info@DeFlockILM.org. Last updated October 5, 2026.