The Pole Is Not the Product

Flock did not sell Texas a camera. It leased cities a searchable graph of other people’s cars.

People keep arguing about a black box on a pole. Police chiefs hold up recovered trucks and missing children. Opponents hold up 1984. City managers hold up a vendor invoice and a transparency portal. All three groups talk about a camera because a camera photographs well and fits on a council agenda.

Here is the thing the agenda never quite names. Flock Safety did not become a national fixture by selling poles. It sold a searchable graph of vehicle movements, leased by the year, federated across thousands of agencies, with software that can find a car when nobody types a plate. Until a city names that object, every meeting about “the cameras” talks past the contract.

Two pictures dominate the public mind. Both miss.

The friendly picture treats Flock as a modern cousin of the red-light camera, minus the ticket. A unit sits at an intersection, reads plates, pings an officer when a stolen car or a missing person rolls past, and forgets the rest in thirty days. No face. No name. Local cops only. If you do not commit a crime, you have nothing to worry about.

The hostile picture treats every pole as an all-seeing eye that follows you from church to the clinic to the school pickup line, feeds a forever file in an Atlanta cloud, and hands the federal government a ready-made map of your week. Some of that picture is fear filling gaps. Some of it tracks what the logs show once anyone bothers to read them.

The live system sits between those cartoons, and it is worse in a specific way. It does not need your name to reconstruct your life. It needs your car.

What Flock actually is

Flock Group Inc., doing business as Flock Safety, is an Atlanta company that three Georgia Tech alumni started in 2017 after a property crime left founder Garrett Langley with no evidence to hand police. The first unit was improvised. The company now describes itself as a public safety operating system. That phrase is not decoration. It is the business.

The hardware is the on-ramp. Pole-mounted automated license plate readers photograph the rear of passing cars. Solar panels and cellular links put them on a right-of-way in an afternoon. Flock now talks in terms of more than a hundred thousand devices across forty-nine states and more than five thousand communities, with tens of billions of vehicle reads a month. Texas alone accounts for hundreds of agencies. The customer list does not stop at police departments. Homeowner associations, retailers, schools, churches, and apartment operators buy in, then many of them share with police. That is how a “city camera program” becomes a sensor net the city does not fully map.

The camera does not capture “a plate.” It captures a plate plus a vehicle signature: make, color, body type, and distinguishing features such as a roof rack, a bumper sticker, or a dent. Time and location tag the hit. The file goes to Flock’s cloud. For years the brochure promised a thirty-day rolling delete. The company now talks new customers toward seven days. The order form and local law set the real number. Hits tied to a case can live far longer. Some large departments have kept investigative copies well past the brochure.

Flock’s public line is consistent and, on its own terms, often true. The company says it does not sell the data. The customer owns it. The ALPR units do not perform facial recognition. The system does not look up your name, address, or Social Security number. Private customers do not receive law-enforcement data. Every search creates a log. Those statements answer a charge the product was never built to commit. They do not answer the charge the product was built to perform.

Three layers sit on top of the pole. Confuse them and you buy the wrong thing.

Capture

Capture is the photograph. A car on a public road is not a secret. Toll roads and parking garages have read plates for decades. If Flock were only this, the fight would concern street furniture and a short-term locker.

Analytics

Analytics is why departments sign. Officers do not only type a plate. They search by description and window: silver Honda, after 9 p.m., these intersections. They run convoy detection, plate-swap flags, partial plates, and visual search. A car with no readable plate is not invisible if the software still knows the dent and the ladder rack. That is not CCTV a detective rewinds. That is a reconstructed trip. The same query box that finds a stolen truck can find two cars belonging to an officer’s child’s mother. Kentucky proved it this week.

Network

Network is the business. Flock’s own site sells coast-to-coast reach: thousands of agencies, shared hotlists, national crime files, statewide lookup, national lookup on a full plate, one-to-one partner sharing, and radius sharing. The value of six cameras in Wylie was never six cameras. It was six cameras plus everyone else who joined the pool. Police chiefs are not wrong that a suspect does not stop at the city limit. They are incomplete when they describe that fact as “our cameras.”

The problem is not the pole

A city camera is a budget line and a public meeting. A national lookup is a checkbox that makes the first camera more valuable the day the second city joins. Once that graph exists, every agency with a login, a hotlist, or a “statewide sharing” toggle becomes a consumer of everyone else’s streets.

The claimants arrive in a predictable order. The home department comes first, for stolen cars and homicides. That is the slide deck. Neighboring departments come next, because crime ignores city limits. Statewide fusion centers and the Department of Public Safety follow, because Texas already thinks in statewide systems. Out-of-state agencies get swept into a “Texas” pool when someone miscategorizes an originating agency. Federal consumers arrive directly or through a fusion center. Then comes the officer with a messy personal life and a purpose-code field that accepts the word “narcotics.” The city council arrives last, after a resident files a records request.

You do not get to pick only the fingers you like. That is the entire argument. Swap Flock for Axon or Motorola and you change the crust if the replacement can still federate. Take the poles down and you revoke collection. Leave the poles up and “tighten sharing” and you place a bet on a setting nobody certified this week.

Kentucky: the query box

Authorities in Shively, Kentucky, arrested Officer Asad Zahir after Flock’s new auditing tool flagged 2,048 searches on two vehicles registered to the mother of his child. He ran those searches from January into May. Investigators counted 241 of them after she obtained a protection order. He labeled the queries a drug or narcotics investigation. Police say she was not the subject of any such case. He had been with the department since 2022 and worked patrol.

Shively enabled the vendor’s AI audit feature on August 24. The tool flagged the pattern at once. By Tuesday they had him in custody on misdemeanor counts of official misconduct and unlawful access to a computer. The department suspended him without pay. Chief Andre Bottoms called the case a gut punch, promised a policy review, and delivered the standard defense: the cameras work when people use them properly, and the public should not judge the tool by one officer.

The tool did not catch a genius. It caught an officer who searched the same two plates more than two thousand times and typed “drugs” in the reason field. Any competent access review should have seen that without a new product. Flock spent years treating anomaly detection as an optional extra. The company moved toward mandatory case numbers and automatic review after the Washington Post and civil-liberties researchers documented dozens of officers who used plate-reader networks to track wives, girlfriends, exes, and women they wanted to meet. Institute for Justice tallies put romantic-stalking cases in the dozens and total ALPR-abuse incidents well over a hundred. Those counts grow every month, and the people who keep the databases say the public file is the tip of the iceberg.

The pole did not stalk anyone. An unbound purpose field and a multi-state search graph did.

Pflugerville: the pie

Pflugerville is the other failure, and it is the one Texas cities can no longer treat as a blue-city story.

On August 25 the city council voted unanimously to kill the Flock agreement rather than approve a two-year extension. Staff had already bagged the twenty-eight city-leased cameras on the city manager’s order. Pflugerville ISD pulled its campus units the next morning. Police Chief Willie Richards said the city had lost confidence in Flock’s ability to meet its data and privacy standards. Mayor Doug Weiss said the city cannot partner with an organization it cannot trust.

A resident, Sam Aly, forced the issue. He used an open-records request to pull audit logs covering February 16 through August 15. Councilmember Jonathan Coffman recited the punchline in open session. The city believed it had authorized about eighty partner agencies, mostly Texas law enforcement. The logs showed 459 agencies in the data. Outside users ran about 1.6 million searches against Pflugerville’s cameras. Local police ran 5,078. That ratio is roughly 315 to 1.

Read the 1.6 million correctly. Those are search events that hit Pflugerville camera data, not 1.6 million residents individually hunted. A single stolen-car plate queried by two hundred agencies inflates the count. The ratio is still the story. When outside agencies out-search the home department 315 to 1, the cameras are no longer a local investigative tool. They are a node on someone else’s network.

Coffman said the quiet part. The city believed it had excluded itself from statewide and nationwide databases. The audit said otherwise. Flock answered that Pflugerville had participated in Texas’ statewide sharing network since 2024, that statewide sharing differs from one-to-one partner sharing, that agencies receive monthly emails about their sharing relationships, and that two out-of-state agencies had been miscategorized as Texas agencies and swept into the pool. If that account is accurate, the city failed to watch a setting in daylight for two years. If the city’s account is accurate, the configuration model cannot be trusted. Either way, the authorized-partner list and the actual search graph were not the same object.

Pflugerville had a transparency page. The page did not show 459 agencies. The resident had to pull the file the portal was not built to explain.

Tempe: leave the category

Tempe, Arizona, announced on August 26 that it had shut the system off, canceled the Flock contract, and would not hire a replacement license-plate-reader vendor. That is a full exit from the category, not a logo change. The city had already stopped sharing plate data with other agencies and stood up a transparency portal. The portal dies with the program. Tempe said its own audits found no local misuse. Mayor Corey Woods still killed it. The cameras helped investigations. The risk of misuse was too great.

Cave Creek terminated the same day. Surprise, Chandler, and Pinal County were already moving. Chandler’s exit followed officer misuse. Pinal’s sheriff pointed to Fourth Amendment concerns. Arizona’s attorney general opened a statewide review the next day.

Denver charts the other path. That city dropped Flock and signed Axon. Count it as a Flock loss. Do not count it as the end of the graph. Vendor-shopping is how a city keeps the pie and changes the crust. Tempe understood a clean local audit does not bind the next county.

Wylie won a date. Carrollton has not won a vote.

DeFlock in Texas is not a party and not a headquarters. It is a map plus a method. Volunteers photograph poles, file public-information requests, read the auto-renewal clause, put a petition in front of council, and show up on the night the contract would otherwise renew itself. The coalition is mixed on purpose, which is why it travels here. Liberals worry about a surveillance state and federal consumers. Conservatives worry about a private company building a warrantless movement graph. Parents worry about the pole on the walk to middle school. They agree on the contract.

Wylie is the cleanest North Texas win so far. The city installed six Flock cameras in August 2024 at $27,000 a year and paid for them with asset-forfeiture funds. Procurement was sole-source. Approval rode a consent agenda. The current term runs through October 13, 2026, and the contract auto-renews for twenty-four months unless the city gives written notice at least thirty days out. Miss September 13 and Wylie locks in through October 2028.

Austin Gregory built DeFlock Wylie around that calendar. Residents pulled the contract, the sole-source letter, the policy, the minutes, and the logs under the Texas Public Information Act. The petition cleared seven hundred signatures. On August 25 the renewal was item one. Assistant Chief Tommy Walters answered with casework, including a murder earlier this year. Mayor Matthew Porter split the difference in one sentence the rest of Texas should steal. The cameras are a good tool when officers are actually solving cases. His concern is that the system stores the data and shares it with a lot of other departments. The council voted 7-0 not to auto-renew.

Two caveats, because honesty is part of the method. The council did not ban Flock. It declined the automatic renewal and left a work session open for a possible future contract with tighter terms. The cameras stay up until mid-October. A later council can put them back. What residents won is time, a public process, and the end of a quiet auto-renew. Most of these systems die, when they die, on a date nobody was watching.

Carrollton is the other side of the same metro. Neighbors noticed poles on residential streets and outside a recreation center across from the community pool. Hannah Foust is the name that keeps showing up in group chats, council comment, and local radio. The Change.org petition to end the program is past twelve hundred verified signatures. The city has not moved.

The camera count is messy because the city and the press use different denominators. Mayor Steve Babick told reporters the city had fifty to sixty Flock cameras on thoroughfares, intersections, and traditional high-crime areas, and that he had no plans to remove them. A city spokesperson later put the figure at sixty-seven. Treat “dozens, not six” as the safe fact. Babick’s defense is operational and vivid. In a thirty-day window he cited 1,710 alerts: forty-three stolen vehicles, seventy-six warrant hits, six missing people. Council meetings now feature both microphones. One resident wants the truck found. Another has counted seven cameras on the commute to work.

Carrollton also did something Wylie did not. In 2025 the police department launched Carrollton Safe with Flock: a voluntary registry of private cameras, plus an option to pipe live doorbell-style feeds into the department. That is a different product from the roadside readers, and it is the same vendor inside the same trust argument. DeFlock Carrollton is fighting a city that already treats Flock as a platform. When the group asked for emails among the police department, the city manager, the mayor, and Flock, the city quoted them $32,000. Wylie residents got the contract file. Carrollton residents got a bill.

Scale and calendar explain the split. Ending Wylie meant “do not sign the renewal.” Ending Carrollton means ripping out a network the police department now treats as infrastructure. The arguments are otherwise the same arguments that just took down Pflugerville: nationwide searchability, partner lists that do not match the logs, officer-misuse cases stacking up in other cities, and original procurement that skipped a real public fight.

Sunlight is a detective control

Transparency is an improvement over the years when these units went in on consent agendas and nobody in the chamber could name the sharing topology. It is not a fix.

A portal tells you, after the fact, that the pie was eaten. It does not keep hands out of the pie. Pflugerville had a transparency page. Wylie had a policy and logs. Tempe built a portal and then killed the program anyway. The residents who moved those councils did not take comfort from dashboards. They used disclosure as a crowbar.

The citizen is not the auditor. A real audit function has time, tools, authority, and a duty to look. A resident has a job and a public-information request that may come back as a five-figure invoice. Sam Aly is the exception the system now pretends is the design. You cannot run accountability for thousands of agencies on the assumption that every suburb contains one person who will parse six months of search events and brief the council before the auto-renew date.

Disclosure also lags use. The query is instant. The portal is monthly, quarterly, or whenever a journalist asks. Zahir ran searches for months, including under a protection order, until an optional vendor module came on. A transparency page published in August does not un-search January. Location history is not a budget line you claw back.

The portals measure the wrong thing. City pages advertise cameras, unique vehicles detected, local search sessions, hotlist hits, and thirty-day retention. That is the collecting agency talking about itself. The citizen problem lives on the other side of the query: who else searched this city’s reads, under what purpose code, from which agency, and whether “statewide sharing” meant what the mayor thought it meant. Have I Been Flocked exists because the event that matters is not “was I photographed.” It is “did someone search me.” Those are different records.

Sunlight can become a permit. “We have a transparency portal” now sits in the sales deck next to “agency-controlled sharing.” Cities that keep the cameras will lean harder on dashboards because dashboards cost less than taking poles down or breaking federation. Transparency becomes the thing you point to so you do not have to change the software.

Citizens are not asking for a nicer chart of their own surveillance. They are asking for a system that cannot be used on them except for a named case by a named agency with a revocation path. That is purpose-binding and least privilege. Transparency is how you verify those controls. It is not the control.

The stack that would match the objection looks like this. Collection can stay local. Search requires a case number the system can validate, not free text. Partner access is an allow-list with an expiration date, not a statewide pool you discover in a monthly email. Out-of-network queries fail by default rather than log and explain later. The collecting city can see, in real time, who is hitting its reads. Logs are public and boring because the denied queries never happened.

Even that will not satisfy everyone, because some residents object to the graph existing at all. Fair enough. It would at least separate “I do not want my car photographed on a public road” from “I do not want another state reconstructing my week.” Those are different complaints. A portal lumps them together and answers neither. It just makes the second complaint provable.

The national wave is real. The network is not dying.

There is no official national roster, which is why the headlines bounce between fifty-six, ninety, one hundred five, and one hundred twenty-eight. The Institute for Justice lists 128 local governments that canceled or declined to renew an automated license-plate-reader contract since the start of 2025. That list excludes pauses and vendor swaps, so a city that leaves Flock for Axon does not count as a win for “no plate readers.” Almost every named system in that database is Flock. A separate tracker, Who Approved This, documents ninety-seven endings as of August 27 and splits them into canceled, expired, rejected, and suspended. DeFlock and the press citing it have used fifty-six cancellations, rejections, or deactivations in 2026 alone, with broader tallies north of one hundred depending on the week. Treat any single number as a floor.

Four different acts get lumped under “they dropped Flock.” A council kills a running contract. A term expires and nobody signs again. A town rejects the cameras before they go up. A city bags the lenses or defunds the line while the hardware stays. Fort Wayne’s council voted 8-0 against $120,250 in funding after residents packed the chamber. Police say thirty-six cameras remain up on a month-to-month hold while they shop a smaller deal. Lansing’s council voted 7-1 to urge the mayor to cancel. That is a resolution, not a dead contract. Asheville voted 5-2 to terminate eleven city cameras and cut software access. Other Flock units inside Asheville city limits that do not sit on the city contract can stay.

This week is not a trickle. Tempe and Cave Creek went on the same day as Pflugerville and Wylie. Kendall County declined to renew more than fifty cameras. Christiansburg, Virginia, voted 6-0 to terminate. Duluth is letting a grant-funded contract die in September, the largest city in Minnesota to peel off. Massachusetts and Rhode Island are stacking non-renewals and pre-deployment kills. Los Angeles let a 138-camera agreement expire after an inspector-general review found a high false-positive rate on stolen-vehicle alerts. A bad stolen-car hit is not a privacy abstraction. It produces a felony stop. Mountain View, Evanston, Oak Park, and Dayton moved after audits showed unexpected federal or out-of-policy search traffic. Austin let a contract lapse. Hays County became the first Texas county to pull the units. Bandera, a Hill Country town of about nine hundred people, terminated. Guadalupe County paid an early-termination fee to get out.

Four motives keep repeating. Cities discover sharing they did not think they approved. Officer misuse in some other jurisdiction becomes impossible to ignore. Error produces a gun in the street. Process fails: sole-source consent agendas, auto-renew traps, unread monthly emails, records invoices that cost more than the cameras. Cost shows up in the smaller non-renewals. Vandalism shows up in Minnesota and in scattered felony cases. Neither is the main driver of the 2026 wave. The main driver is the discovery that the partner list and the search graph are different objects.

Do not write a eulogy. Flock still talks in terms of five thousand-plus communities and tens of billions of reads a month. Dallas sits on a multi-year contract worth up to $5.74 million. Carrollton’s mayor is holding. Most large Sun Belt departments have not moved. Exiting Flock has become a routine municipal act. The installed base still dwarfs the walk-aways. Both facts can sit in the same paragraph.

What to take to council

Ask questions that force a document onto the table. “Do you care about privacy?” dies in the chamber. “Show us the sharing graph for the last six months” does not. The point is to separate the pole from the network, and to make a council answer with a setting, a date, and a name.

Ask what the city actually bought.

What contract is in force, with what vendor, what term, what annual cost, and what fund pays for it: general fund, grant, or forfeiture? Did the city bid it, or did a sole-source consent agenda quietly approve it? How many devices sit on the city contract, and how many other Flock or plate-reader units sit inside city limits on school, county, highway, HOA, or private contracts the city does not control? Does the contract auto-renew? What is the written-notice deadline, and who owns that calendar? Which software modules did the city purchase: plate read only, description search, convoy detection, statewide or national lookup, audio, live video, private-camera registry?

Ask who can query the city’s reads.

Produce the current allow-list of agencies authorized to search data from these cameras. Produce the audit of who actually searched them for the last six or twelve months: agency name, search volume, and whether each agency sits on the allow-list. How many of those searches came from the home department versus everyone else? Which sharing modes are on right now: one-to-one partners, radius share, statewide lookup, national lookup, shared hotlists, fusion-center access? When did a named city official last certify each mode, rather than the vendor? Can the city see, in the console, a live list of outside agencies hitting its cameras, or does that fact arrive only after a records request?

Ask what “we turned off nationwide sharing” means.

Statewide lookup and national lookup are different switches. Hotlist alerting still feeds national stolen-vehicle and missing-person files even when “lookup” is off. Has any out-of-state agency been grouped as in-state by mistake? Who receives the vendor’s monthly sharing emails, and are those emails in the official record? If staff says sharing is agency controlled, ask whether they mean the contract language or the admin panel this week.

Ask what binds a search to a real case.

Does the system require a validated case number, or will it accept free text? Who reviews purpose codes, how often, and when did the last review happen? Has the vendor’s anomaly tool been enabled, on what date, and who receives the alerts? In the last year, how many officers in this department were flagged, suspended, or charged for plate-reader misuse? Does the city audit outside agencies that search its data, or only its own users? “We trust our people” is not a control.

Ask what the system collects, keeps, and can delete.

What fields sit beside the plate? What retention period is in the order form, not in the press release? When a hit is saved to a case, where does that copy live, and who can still query it after the rolling delete? Does “customer owned” mean the city can export and destroy the data, or that the city has a login? Has the city ever demanded wholesale deletion, and did the vendor confirm destruction in writing?

Ask what happens when the system is wrong.

What is the false-hit rate on stolen-vehicle and warrant alerts? How many stops came from alerts that turned out to be misreads? Does a supervisor review a hit before a felony stop, or do officers roll up hot? Los Angeles walked in part because bad stolen-car alerts put guns in the street. Ask whether this department has counted that risk at all.

Ask how the program got approved and how it ends.

Which ordinance or vote authorized it? If the answer is a consent agenda, put that on the record. Is there a surveillance-technology ordinance that requires a public hearing before renewal or expansion? What does mid-term exit cost versus notice-window exit? If the city terminates, do school-district, county, and HOA cameras inside the city stay live and searchable by others? Will the council leave the category, as Tempe did, or only the vendor? Asheville’s mayor had to kill the contract because a moratorium would not let the city remove the hardware. Ask whether this contract has the same trap.

Ask what a resident can obtain without a lawyer.

What is the fee estimate for the audit logs, the sharing-settings screenshot, the contract, and the vendor emails? Make the council hear that number in public. Will the city publish a redacted search log on a schedule, including outside-agency volume, without waiting for a request? Is there a named administrator whose job includes recertifying partners every quarter?

Demand one packet before the next renewal or expansion vote: the contract and order form; the auto-renew date; a screenshot of current sharing toggles; the authorized-agency list; six months of search volume by agency; the purpose-code rules; the last internal audit; the retention settings; and a staff memo that answers whether statewide and national lookup are on. If staff cannot produce that packet, the city is not running a camera program. It is renting a network it has not looked at.

Skip the sermons. “Isn’t this 1984,” “don’t you want to catch criminals,” and “will you promise no one will ever misuse this” waste the three minutes. The first two are hymns. The third is a promise no council can keep. The questions above are designed so that “we will get back to you” is itself an answer. They do not know who is in the pie.

Call the thing by its name. A Flock deployment is a subscription to a multi-agency vehicle-intelligence graph, with cameras as the collection method. Argue about that. The pole is just where the argument got stuck.