The Column Disappeared. So Did the Explanation.

This blog and website have raised quite a few questions and left them unanswered, like “what is the search/lookup inversion that happened in August?” and “what were the 250,000+ cameras seen in a search from Missouri?” Flock may have cleared up the mystery at least a little bit in a recent email: Flock appears to be working to unify its network.

Whether that is already done or will be done soon is unclear, but it answers another recent question: “what happened to the ‘devices searched’ column?” A few weeks ago^[1] that column disappeared. We know Flock likes to alter contractual terms and unilaterally remove audit capabilities. We also know that its “Devices searched” field disappeared for a while between November 2024 and February 2025.^[2] It has been unclear if this latest removal was on purpose, and, if so, what that purpose might have been.

A quasi-enlightening email exchange^[3] begins in August 2025, when a police officer asks Flock about “vehicle reads from the Condor cameras.” (Quick reminder: Condor is Flock’s PTZ “AI-powered video” camera that automatically zooms in on your phone). This was confusing to the officer because Condor cameras are not ALPR devices. He, like many elected officials and police officers, likely did not understand that Flock’s LPRs aren’t either.

There is no meaningful difference between an “LPR” (which, by the way, is Flock’s product name, not its category) and a “Condor.” They record things, they are analyzed via machine learning or AI, or Upwork contracts, and they are searchable by anyone Flock chooses to give access. Both the “LPR” (formerly “Falcon” and “Sparrow”) and the “Condor” are AI-powered surveillance cameras.

The officer’s inquiry came about 4 months before YouTuber Benn Jordan disclosed a separate security failure on the same Condor line: debug interfaces on what Flock characterizes as “a very small number” of units had been left Internet-accessible without password protection. Flock said it was a limited, isolated configuration issue and blamed it on Verizon. It was, however, another symptom of having no organizational controls in place. Both incidents involve Condor cameras doing things they weren’t supposed to do, and Flock did not proactively notify its customers about either.

A little over a month after the officer’s email, a search revealed >257,000 cameras. The log entry was accompanied by a note:

An issue was identified that caused the system to initiate unprocessed search activity on a larger set of cameras than intended by the user. No footage or data from these devices was accessed or viewed. The underlying bug has been fixed, and additional safeguards have been implemented to prevent recurrence.

At the time, I noted that the note does not deny that the cameras were real. I concluded that the “number includes Flock’s retail customers, like Lowe’s, Academy Sports, and FedEx.” My conclusion was based on an assumption that Flock would not merge what it holds out to be its “LPR” network and its general-purpose “AI-powered surveillance” network. I now believe that assumption was wrong.

Flock checked with engineering and responded to the officer’s email. The response was that Condors should not be showing in the “vehicle reads,” but that the engineering team was working on some new features for 2026. It was “refining” some of the data. That, according to Flock, is what caused the issue. The answer was not any more specific.

What “refining data” means is anyone’s guess, but what we do know is that Flock was rolling out FreeForm around that time. FreeForm went into “Early Access” in March of 2025. “Early Access” is a term you’d expect to find on Steam, not in your mass surveillance contract. On Steam, the term is widely understood to mean “buggy and not suitable for any serious use” — a warning that your save may be deleted or your PC might crash. Flock slaps the label on a technology that can search for people, where it’s constrained only by a broken AI moderator.

I’ve often mentioned the other oddities in August 2025; possibly first in an analysis of logs from Santa Cruz, where 90% of logs disappeared overnight. That data also showed a massive change in “search” vs. “lookup” usage, a pattern that would become visible across organizations. For example, Houston, TX, did 492,000 searches in May but “only” 183,000 in August (still roughly 6,000 searches per day).

Maybe by August our civil rights were ready for beta.

These anomalies have never been mentioned in an audit report that I’m aware of, and Flock has not addressed them other than maybe through this statement about data refinement.

After the email exchange, and reviewing data from the FreeForm report, I now think it’s likely that the 250,000 number reflects Flock’s broader network — with or without retail customers — including Condor and potentially other AI-powered surveillance cameras. The email doesn’t say that. But Flock was actively “refining data” across camera types at the same time FreeForm soft-launched, and the 250,000-camera anomaly appeared one month later. That’s circumstantial. It’s also the most coherent explanation available until Flock, or any of its customers, offer a better one.

After unification, maybe Flock will (or does) offer a search interface that attempts to maintain illusory ALPR status, or maybe it will (or does) have a single interface for searching over a quarter of a million cameras that automatically follow you around and zoom in on your phone. Its AI moderator could continue to approve terms like “jeans” and “tweaker on a bike”, or it could not.

Maybe Flock will bring back the “Device count” column some day, maybe not.

We still don’t know what they’re refining or why they’re refining it, but whatever it is, they’re doing it in secret. The column is gone, the data changed overnight, and not a single audit report has addressed it.