The City Council Study Session Is a PR Campaign for Flock.

The city's own search logs say otherwise. The Casa Grande PD account of Brandon Hill ran 42 searches between January 3 and March 3, 2025 with reasons like "target veh USBP," "target veh ERO," "ERO FUG OPS," and "USMS/ERO," plus 5 more "load vehicle" searches with the tribal police on the border. Some covered more than 300 camera networks and 5,000 cameras.

This is the pattern 404 Media documented in May 2025: ICE has no Flock contract, so local officers run the searches for it. We do not know the arrangement here, and these searches happened before the September 2025 contract. The handout speaks in the present tense and says "No."

Even if it stopped tomorrow, a city policy cannot bind the federal government. A subpoena compels Flock directly, and the contract lets Flock hand footage to "law enforcement authorities, government officials, or third parties."

Officers can, and did. Flock sells a search box called FreeForm, and Casa Grande has it turned on. From October 2025 through January 2026 the logs show 84 searches for people by description. Stephanie Sjerven typed "male." Damian Guerrero typed "female wearing grey sweater." Kyle Short typed "person wearing black hoodie and black hat." Jeffrey McKean typed "child."

On December 9, 2025, Damian Guerrero typed "native female wearing grey hoodie with green lettering and black pants." Flock blocked it. Thirteen seconds later he typed the same thing without "native." It went through.

On October 21, 2025, Caleb Zadareky typed "hispanic male with a grey shirt, beanie, blue jeans and a backpack." Blocked. Later that day Christopher Mendez ran the same description without "hispanic" across 353 networks. Allowed.

Many of these searches were for runaways and missing people. That may be good police work. It is still a capability the council was told does not exist, and Flock's own patent describes sorting people by "male, female, race," clothing, and "height and weight."

On January 30, 2026, a deputy chief emailed command staff a list of "notable wins" from the camera system's first ten days of live monitoring. One of them: "Monitoring of a student protest at City Hall."

The same email says the officer watching the cameras can "identify involved parties" and "monitor activity prior to patrol arrival," and that officers used the cameras to spot "suspicious activity, primarily in park areas."

The department's written policy only bans monitoring people "based solely on" politics. Watching a protest for "public safety" slips right through that word. The handout promises more than the policy does.

Read the list again. Plate cameras, video cameras, drones. The 31 gunshot-detection devices are missing from it. Those are microphones.

One page earlier, the handout says the gunshot system detects "gunfire and other critical public safety events." A microphone cannot tell a gunshot from a scream without listening to everything.

Flock has announced that these microphones will start listening for human "distress" sounds.

The logs show Casa Grande officers searching far beyond Arizona. Of 18,413 searches in 13 months, 5,406 reached 100 or more camera networks and 62 reached 1,000 or more. One theft search in late November 2025 covered 6,504 networks and 97,758 cameras. That is the whole country.

Sharing on Flock normally runs both ways. If Casa Grande can search other agencies' cameras, ask which agencies can search ours. The handout does not say.

169 searches were labeled "Agency Assist." The handout does not say for whom.

Every layer in the handout is the department checking itself. Sergeants check officers. Lieutenants check sergeants. The "Audit Assistance Tool" is Flock's software checking Flock's customers. The Chief decides. There is no outside audit, and the handout does not promise one.

The department's own policy says every search needs a case number and a reason. We read all 18,391 searches in the released logs. 10,954 have no case number at all: blank, "none," a word, or a few digits. Another 1,452 use a placeholder like "2025-00000000." That leaves 12,406 searches, two out of three, with no real case number. We counted anything that looked like one, even a five-digit number.

The reasons are not much better. 2,791 searches give no reason. 7,446 more give only a word: "ATL" (3,194 times), "investigation" (2,522), "suspicious" or "suspect" (420), "test" or "training" (99). Put together, 7,089 searches, about 4 in 10, have no real case number and no real reason. The border and immigration searches mostly carry one code, "4Z15BR254Z0001," or nothing.

The audit only records searches. It does not record an officer watching live video, so monitoring a protest would never show up in it. And when Flock flagged two searches for race, the policy says they go up the chain to the Chief. The handout does not say whether that happened.

The cameras photograph every car that passes, whether or not anyone is suspected of anything, and keep the record for 30 days. That is the definition of general surveillance. Only the search is targeted.

The handout says the goal is "citywide coverage." It also admits investigators check whether a car "was in a particular area during a relevant timeframe." That kind of search returns every innocent car that drove by.

The plan is about 240 devices for about 60,000 people, including 100 video cameras that record all the time. That is roughly one device for every 250 residents.

That describes Flock's software. It does not describe what an officer can do with a picture. Nothing in the handout or the department's policy stops an officer from taking a face from a camera image and running it through another tool, like Clearview AI.

Every "do not" on that list is about the plate cameras. The 100 video cameras "continually record activity in their field of view." That is the handout's own sentence, and the people searches above ran against those video cameras. Whenever a promise says "ALPR," ask what the same promise looks like for the video cameras, the drones, and the microphones.

Notice the words "independently" and "for commercial purposes." Now read the contract.

The city gives Flock a "non-exclusive, royalty-free, irrevocable, worldwide license" to use the data. Flock may also "collect, analyze, anonymize" it and use it to "train machine learning algorithms," with a "worldwide, perpetual, royalty-free" right to what it learns.

Not sold is not the same as not used. Casa Grande pays $10 million and hands Flock the raw material for its AI for free. Google pays for the imagery it collects. Flock gets Casa Grande's from taxpayers.

The handout gives no numbers. The department has them. The same January 30 email attached a table of every arrest from a camera hit since December 2023.

The total is 73 arrests in 26 months. That is about 3 a month. 35 were car or property crimes, 21 were drug cases, 11 were warrants, 2 were weapons cases. Four were violent crimes. Three months had none.

The program costs about $1 million a year, about 13 police officer salaries. In 2025 alone, officers ran about 17,500 searches. The department counted 23 arrests from camera hits that year.

The biggest Flock case in the country is not over. Two Norfolk, Virginia residents sued over 176 Flock cameras. In February 2025 the court said their claim of dragnet surveillance was plausible enough to go forward. In January 2026 the judge ruled for the city. They appealed. The appeal is pending right now.

Read the judge's reason. He said 176 cameras did not capture "the whole" of a person's movements. Norfolk has about 240,000 people. Casa Grande is planning about 240 devices for about 60,000 people. That is roughly four times the cameras per person, plus 100 video cameras that never stop recording, plus drones.

The Supreme Court has already said that tracking a car with a GPS device for 28 days is a search, and that pulling months of phone location history is a search. Nobody has ruled on a network this dense. "Not yet ruled unconstitutional" is not "constitutional."

You do not need radar to measure speed. Every camera knows its location and stamps the exact time on every photo. Two cameras on the same road give distance and time. Distance divided by time is speed. That is how average-speed enforcement works.

Flock's own FAQ says its cameras "do not currently measure vehicle speed." Currently. Flock already sells trailers that combine plate reading with speed feedback.

"Do not issue citations" is not "not used for traffic enforcement." Local reporting described a two-day traffic crackdown using Safe City camera and drone data. The January 2026 log has searches labeled "Traffic Infraction" with the case number "none." The January 30 email says footage was used to determine the cause of traffic collisions.