TL;DR: Joseph Cox at 404 Media published an investigation on July 16, 2026, showing that Flock Safety's FreeForm search feature, marketed as an investigative aid for license plate reader footage, has been used "at least hundreds of times" by police to find human beings instead of vehicles. Audit logs reviewed by 404 Media show officers typing natural-language descriptions of people ("heavy-set male with a black and white hat," "non caucasion [sic] male wearing blue shirt blue pants white hat," "white jeep with trump flag") and pulling footage across hundreds of cameras in a single query. Flock says FreeForm is not facial recognition and blocks searches for race, ethnicity, religion, and nationality. The audit logs show those exact categories in the search prompts. The reporting lands inside our existing Flock cancel-wave cluster, and the same audit log channel is what EFF used to find 80+ agencies searching for Romani stereotypes last year.
What FreeForm Actually Does
FreeForm launched in February 2025. Flock describes it as an AI-powered natural-language search across the company's combined network of automated license plate reader cameras and its newer Condor video cameras [1]. An officer types a description into a search bar and the system's image recognition scans the requested footage for matches, returning a curated set of clips the algorithm believes are relevant.
Flock consultant Kevin Cox, a former Grand Prairie, Texas police officer, told 404 Media that the combined video and plate reader evidence is "an incredibly game changing experience" for investigators [1]. The marketing pitch to cities has consistently been stolen cars and missing children. The actual use case in the records 404 Media obtained is broader: searches by clothing, tattoos, body type, race, gender, perceived political affiliation, and activity [1][2].
The Query Language Is Already People, Not Plates
The audit logs 404 Media reviewed include a "text_prompt" field that captures the search description. The fields read like suspect write-ups, not plate numbers. Examples pulled directly from the records include "heavy-set male with a black and white hat," "person on skateboard," "person wearing orange vest and construction hat," "black sweatshirt," "blue jeans, boots with white paint stains, possibly carrying a black helmet," "female with ugg boots," and "a male on foot" [1]. One description is a near-racial profiling template: "white male about 6ft 1in tall, longer brown hair" [1].
Wikipedia, citing the same 404 Media reporting, summarizes the audit log evidence bluntly: "Public records requests showed that police were searching for characteristics of individuals, including by race, gender, clothing, height, tattoos, and perceived political affiliation" [2]. One entry logged a "white jeep with trump flag" as the search description [1].
This is a different category from the searches the cancel-wave stories have documented. Our earlier coverage of cities tearing out Flock cameras focused on plate reads that ended up in federal hands, including ICE, ATF, Air Force, and GSA Inspector General queries [3]. FreeForm adds a second, more invasive channel on top of the plate reads: the same network that flags your car can now be told to find a person who looked like you.
The Network Is Built for Cross-Jurisdiction People Searches
FreeForm queries do not stay inside one department. The records 404 Media reviewed show officers pulling footage across dozens of networks in a single search. Dunwoody, Georgia police ran queries across 9 Flock networks. Pocatello, Idaho police ran searches against 38 cameras. Milford, Connecticut used "more than a hundred cameras" for individual queries. The California Highway Patrol ran a single gray-shirt search across 274 cameras. The Texas Department of Public Safety ran searches against 96 networks. Anne Arundel County, Maryland police pulled footage across 198 networks in a single query [1].
This is the structural risk the cancel-wave stories have been pointing at: the value of the Flock network to a single department is the rest of the network. The same feature that lets a small-town police department reach across state lines to find a stolen plate also lets it reach across state lines to find a person who fits a description. Once the camera data is in Flock's cloud, the geographic boundary the city thought it bought disappears.
What Flock Says
In a statement to 404 Media, Flock said FreeForm "is designed to help investigators quickly search through large amounts of footage" and "is not facial recognition" [1]. The company told 404 Media it does not develop facial recognition. Flock also claims FreeForm blocks queries for "race, ethnicity, religion, nationality" and sends an alert to the agency administrator when a scrutinized query triggers that filter [1].
The audit log evidence reviewed by 404 Media suggests those guardrails fire constantly. The descriptions captured in the records include explicit racial and political-affiliation prompts. Flock did not say what happens to an officer after the admin alert fires, and the records show the searches still ran. The natural-language interface does not require facial recognition to enable people-tracking: a description of clothing, body type, race, gender, height, tattoos, and political belief is enough for a human reviewer to identify a candidate, and that is the workflow the audit logs document [1][2].
What Civil Liberties Groups Are Saying
Jay Stanley, a senior policy analyst with the ACLU's Speech, Privacy, and Technology Project, told 404 Media the public has not caught up to how much more powerful a surveillance camera has become, and that Flock is moving beyond its original pitch of catching stolen cars toward general-purpose people-tracking [1]. Tom Bowman, policy counsel for surveillance and security at the Center for Democracy and Technology, called the practice a bait-and-switch in which cities bought a stolen-car tool and got a people-tracking system [1]. Beryl Lipton, a senior investigative researcher at the Electronic Frontier Foundation, said AI-enabled video analysis across large volumes of footage creates new risks for people going about their daily lives [1].
EFF has separately used the same Flock audit log channel to document 80+ law enforcement agencies running searches that targeted Romani and Irish Traveller communities with stereotypes [4]. The FreeForm disclosures and the EFF Romani-stereotype audit are two windows into the same database.
What to Watch
The audit log records are the part to watch. FreeForm queries are logged with a "text_prompt" field in the data Flock itself produces, which means every camera-equipped department already has the receipts sitting in its own system. Cities debating Flock renewals should be filing the same public records requests 404 Media used. Watch the California Highway Patrol, Texas Department of Public Safety, and Anne Arundel County specifically, since those three departments ran the largest documented FreeForm queries [1].
Watch LAPD's contract expiration, finalized July 11, 2026, for whether the Board of Police Commissioners takes up the FreeForm question alongside the 161-stop audit that ended the contract. Watch the cancel-wave cities for whether any of them cite FreeForm by name as a renewal-blocker. And watch whether any state legislature moves to ban the people-tracking use case specifically while allowing the license plate reader use case to continue, since that is the policy distinction the Flock marketing pitch has been trying to keep blurry.
The cameras are still up in most of the country. The FreeForm disclosures turn the audit-log demand from a privacy-advocate request into a baseline public-records request any city council member can make tonight.
Sources
- 404 Media: "How Cops Use Flock to Track People, Not Cars" by Joseph Cox (July 16, 2026)
- Wikipedia: Flock Safety entry on FreeForm search feature and audit-log evidence
- State of Surveillance: "30+ Cities Have Canceled Flock Safety Contracts" (May 19, 2026)
- State of Surveillance: "DeFlock and the 90,000-Camera Revolt" (April 2026)