The Camera That Remembers Every Car
Published
A Camera on Every Corner
Drive through a lot of American suburbs today and you will likely pass one of these without noticing it: a slim, solar-paneled camera bolted to a pole at a neighborhood entrance, angled at the road. It isn't a traffic light or a speed camera. It's a Flock Safety automated license-plate reader (ALPR), one of a rapidly growing number quietly cataloging the comings and goings of every vehicle that passes beneath it. Flock Safety, an Atlanta-based company founded in 2017, has built one of the largest privately operated vehicle-surveillance networks in the country by selling small-town police departments, homeowners' associations, and business districts a subscription camera service marketed less as "surveillance" and more as a modern neighborhood watch. The pitch works because it's true, up to a point — these cameras really do help recover stolen cars and close out break-in cases. What gets less attention is how much a plate-reading network can see once it's networked at scale, who else gets to search it, and what happens to that record once your neighborhood decides to plug in.
How the Cameras Work
Each Flock camera is a largely self-contained unit: solar-powered, battery-backed, and connected over cellular data rather than wired into fixed power or a local network, which is part of why a camera can go up on almost any pole in a single day. Motion in its field of view — a passing car — triggers a still-image capture. From there, the analysis happens through Flock's own computer-vision pipeline rather than a person watching a monitor: software reads the plate text, then extracts a set of visual attributes about the vehicle itself, such as its make, body type, color, and distinguishing features like roof racks, bumper stickers, or visible damage. Flock markets this second capability as a "vehicle fingerprint," and it may be the more consequential one — it lets an investigator search for something like "silver pickup with a ladder rack and a dented rear bumper" even with no plate number at all, a query a purely plate-based system could never answer. Every detection — plate, fingerprint, timestamp, and location — is uploaded to Flock's cloud platform, checked against hotlists for stolen vehicles or active alerts, and retained for a period the company says defaults to about 30 days, configurable by the customer, before automatic deletion. No one is watching a live feed; the system runs unattended and simply logs a record of every plate that passes, whether or not it is ever relevant to anything.
Who's Watching
What makes Flock notable isn't the hardware — automated plate readers have existed for two decades — but the pace and breadth of its rollout. The company says its cameras now operate in thousands of communities across the United States, and by mid-2026 the total camera count is somewhere in the high tens of thousands to low hundred-thousands nationwide — Flock itself claims roughly 120,000 cameras, while the independently crowdsourced DeFlock map counted about 96,000 in July 2026, a gap that shows the precise figure is still contested and likely stale within months. Customers aren't limited to police departments: Flock sells directly to homeowners' associations and private business districts, so a meaningful share of the network sits on private property, governed by private contracts, with search access potentially resting with a security guard or an HOA board rather than a sworn officer. The company has also expanded well beyond license plates, adding gunshot-detection audio sensors and drone programs pitched to departments as a faster, cheaper alternative to dispatching a patrol car. Its growth has been rewarded by investors — reporting on Flock's funding rounds has described valuations in the billions of dollars — a scale that reflects how quickly this category of always-on, privately run surveillance infrastructure has become normalized, often with far less public procurement debate than a city would face buying, say, a new fleet of patrol cars.
What the Data Shows
The feature that concerns researchers most isn't any single camera — it's the network. Flock's platform lets one agency opt to share its camera data with others, and many participating police departments do, creating a searchable lookup that can span state lines: an officer in one state can query hits captured by a camera thousands of miles away, in a jurisdiction with entirely different rules about what that search should be used for. Flock requires officers to log a stated reason for each search and has built public-facing "transparency portals" showing search volumes and reasons, a response to years of pressure from civil-liberties advocates. But investigative reporting — notably a series of 2024 stories from 404 Media — found that those logged reasons are often vague or generic, and documented cases in which out-of-state investigators used the nationwide network to run plate searches connected to abortion-related travel and to immigration enforcement, including in places that had adopted local policies meant to limit exactly that kind of cross-agency cooperation. The Electronic Frontier Foundation has tracked Flock's expansion for years as part of its broader Atlas of Surveillance project and has repeatedly warned that plate-reader networks are prone to "mission creep" — deployed to catch car thieves, then quietly repurposed for uses the community that approved them was never asked about. In 2024, the Institute for Justice, a libertarian-leaning legal nonprofit, sued the city of Norfolk, Virginia, arguing that its dense grid of Flock cameras amounted to a warrantless, continuous tracking system in violation of the Fourth Amendment — a test of whether blanket camera coverage of a city is meaningfully different from the kind of persistent location tracking the Supreme Court restricted in Carpenter v. United States.
What's at Stake
None of this requires anyone to have done anything wrong. A plate-reader network doesn't wait for probable cause; it logs everyone, continuously, and only asks whether a particular trip mattered after the fact — if it asks at all. That inversion, collect everything first and justify the search later, is what separates this kind of infrastructure from a single security camera outside a store, and it means the privacy stakes scale with the size of the network in a way that's easy to underestimate one camera at a time. A car that passes five networked cameras a day quietly generates a five-point sketch of someone's movements with no warrant ever sought, and a cross-agency sharing arrangement can make that sketch available to departments the driver has never heard of, for reasons they will likely never learn. Flock's cameras haven't been secretly installed — city councils vote on them, HOA boards approve them, and the company publishes marketing material describing exactly what the system can do. But most people who drive past one have no idea it's there, no way to know how their movements are being logged, and no practical way to opt out short of not driving. Whether that trade-off is worth the crimes it helps solve is, at minimum, a conversation communities are being asked to have with far less information than the system itself collects about them.
This page tracks your reading behavior (mouse/touch movement, scroll, dwell time, clicks) while you're here, and a ~60 second timer runs either way. You can accept or decline. Learn more