Police Are Using Flock's AI Camera Tool to Watch Anyone, Not Just Cars
Wired's analysis of Flock Safety's own browser code shows how its AI search tools work, where the guardrails fail, and why dozens of officers have already turned the system on people they knew personally.
Key points
- Flock Safety's AI search tool lets officers type a plain-English description and have cameras across a defined area run a continuous, automated search for anyone matching it.
- Officers accused of misusing Flock's system have mostly targeted romantic partners or ex-partners, including one who searched a single woman 124 times.
- A Texas deputy used Flock to search for a woman who had had an abortion, prompting two members of Congress to write to the company's CEO.
- Flock's own browser code, reconstructed by Wired, shows its guardrails log misuse but can't stop a determined officer from continuing.
- Political and cultural expression is the only sensitive category where Flock issues a warning rather than blocking the search.
Flock Safety sells camera networks to police departments across the United States. It built its name on licence-plate readers, but its newest tools go further. Officers can type a plain-English sentence, something like "person wearing scrubs," and have every camera inside a chosen area of a city run a continuous, automated search for anyone matching that description.
Wired collected the software code Flock sends to an officer's browser before login, reconstructed what officers see, and examined what the code actually does. That analysis is the basis for what follows. We first reported on officer misuse of Flock's network on 13 August 2026.
How does the search tool actually work?
Flock calls its plain-language search feature FreeForm. Every camera frame gets stored as a set of numbers describing its contents. A typed description, say "blue jacket, red bag," gets converted into numbers the same way. The system compares the two sets and surfaces the closest matches.
A second feature, Smart Sort, lets the officer vote on results: approve an image and similar footage rises; reject it and it drops. A handful of clicks can narrow a large batch down quickly. Hidden from view are any additional filters or instructions running on Flock's servers, which no outside researcher can examine.
Vehicle searches use menus of colours, body types and makes. Person searches offer no such structure. Officers type freely and the AI does the rest.
What are the guardrails, and do they work?
Flock announced changes in August: shorter default data retention, mandatory case codes, plus automated auditing to flag suspicious activity. It says all of these will be required across all departments by year's end.
The code tells a different story. When an officer types a description, Flock's AI scores it against sensitive categories that include race and religion. High scores in most categories stop the search. Political and cultural expression is the exception: the system issues a warning and lets the search continue. Tom Bowman, policy counsel at the Center for Democracy and Technology, told Wired that political expression is "one of the most protective areas" under the First Amendment, calling it a curious place for the lightest check.
The code also shows the guardrails can't block a determined officer. Flock itself warns that results "may be inaccurate or incomplete" but places the "risk of any inaccuracies" on the officer running the search.
Should ordinary people be worried?
Yes, based on what's already happened. Officers accused of misusing the system have mostly used it to track people they knew: romantic partners, ex-partners, people they wanted to contact. A Milwaukee officer searched one woman 124 times, logging each as an "investigation." A Georgia police chief searched his ex-girlfriend and her daughter 600 times before being charged. A North Carolina officer filed 29 of 31 searches for her boyfriend's ex-wife as traffic infractions.
As we reported on 27 August 2026, a patrolman near Atlanta ran 85 searches on two colleagues using the same network.
Experts who study automated surveillance say no AI screening system at this scale is consistently accurate, and nobody outside Flock can measure how often it's wrong or in which direction.
The real problem here isn't a rogue officer or two. It's a system designed to record what officers do after the fact, rather than prevent harm before it happens. Until Flock's server-side controls are open to independent audit, those audit logs are a paper trail that follows abuse rather than stopping it.
Watch for these signs that camera surveillance may be affecting you: unexplained police contact following routine travel, or an officer who seems to know your movements without an obvious reason. If you think you've been tracked without cause, the American Civil Liberties Union has resources for people dealing with police surveillance.
