How Cops Use Flock to Track People, Not Cars
404 Media Joseph Cox
Cops use Flock cameras to hunt for people, not just cars. Vague tags like 'person on skateboard' can scan hundreds of cameras at once.
Based on reporting by 404 Media, Joseph Cox — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Flock's cameras are famous for reading license plates, but data reviewed by 404 Media shows police departments have used the system's AI at least hundreds of times to hunt for people described by little more than their clothes or build. Searches like "heavy-set male with a black and white hat," "person on skateboard," and "person wearing orange vest and construction hat" show up in records. These run through a feature called FreeForm, which lets an officer type a natural-language phrase instead of a plate number and have Flock's image recognition sort through footage the way a search engine sorts text, often querying dozens or even close to a hundred camera networks in one shot.
The examples pile up fast once you start looking. Dunwoody, Georgia searched for someone in a "backpack," then "person walking" and "black sweatshirt" across nine networks each. Pocatello, Idaho ran "a male on foot" through 38 cameras, and separately searched "atlanta falcons." Corona, California looked for an "american flag shirt" and a "dodger shirt." Milford, Connecticut queried "male with tattoos," "male with brown hair" and "woman blue shirt" across more than a hundred cameras. The California Highway Patrol searched 274 cameras for a "gray shirt," Texas's Department of Public Safety searched 96 networks for a man in a black t-shirt and shorts, and Chamblee, Georgia ran an 85-network search for a "white woman wearing grey shirt, blonde hair, black shorts with blue and white shoes," plus a separate hunt for "female with ugg boots."
Some queries go further, naming race outright. The California Highway Patrol searched for a "white male about 6ft 1in tall" with brown hair and paint-stained boots. Atlanta police searched for a "non caucasion male wearing blue shirt blue pants white hat." Anne Arundel County, Maryland ran a 198-network search for a "white jeep with trump flag," which reads less like a stolen-vehicle report and more like tracking someone's politics. Some of these searches are tied to active investigations or missing-persons cases; others have their stated reason redacted entirely.
The people watching this technology closely aren't reassured. Jay Stanley of the ACLU points out that AI video analytics now let police search oceans of footage the same way they'd search a text file, flagging tattoos, bumper stickers, even t-shirts, without needing face recognition to do serious damage. Tom Bowman of the Center for Democracy & Technology calls it a bait-and-switch: towns bought a tool pitched for stolen cars and missing kids, and got something closer to a smoke detector that quietly records every conversation in the house. Beryl Lipton at the Electronic Frontier Foundation raises a darker comparison, noting how many people at any moment are simply walking, wearing a backpack, or have brown hair, and recalling that a hoodie was once treated as grounds for suspicion in the killing of Trayvon Martin.
Flock insists FreeForm isn't facial recognition and can't identify anyone by name, and says it has guardrails blocking searches on race, ethnicity, religion, or nationality, with violations triggering an alert to an agency's administrators. Yet 404 Media found searches that plainly referenced race anyway. The company points to legitimate uses too, including a September 2025 AMBER alert and a case in Emporia, Kansas involving an elderly man who'd wandered from an assisted living facility. FreeForm itself launched in February 2025, though Flock's rollout messaging at the time focused on vehicles, not people.
This isn't the first time Flock's ambitions have outrun its guardrails. Back in May 2025, 404 Media revealed a plan to fold data broker and breach information into a people-lookup tool meant to jump straight "from LPR to person"; Flock dropped it after the reporting and internal pushback. Then in December 2025, 404 Media found at least 60 of Flock's people-tracking Condor cameras left exposed to the open internet, letting anyone watch the feeds live. Each incident points the same direction: a company built around linking data streams keeps finding new streams to link.
My take — AI-written commentary, not fact-checked reporting
Guardrails that let race show up in search queries anyway aren't guardrails, they're a marketing line, and that gap between what Flock promises and what departments actually type into the box is the real story here. A network sold as a way to catch stolen cars was always going to get repurposed the moment natural-language search on people became technically possible, because tools built to combine data streams get combined, policy be damned. The people who should be worried aren't just the ones with tattoos or backpacks fitting some vague description, it's everyone, since the whole point of this system is that anybody can become the search target next.
Read more about this at: 404 Media
