Freedom Beagle crest Freedom Beagle crest Freedom Beagle Off the leash, on the story.

On the Scent

Flock's Trust Page Says the Cameras Don't Watch People. Flock's Product Page Sells People Search.

FREEDOM BEAGLE STAFF · JULY 29, 2026

Flock Safety sells automated license plate readers — pole-mounted cameras that photograph every vehicle passing a fixed point, convert the plate to searchable text, and file it with a timestamp and a location. Its cameras run in thousands of American police agencies, neighborhoods, and businesses, most of them bought with public money and pointed at public roads. For the residents living under those cameras, and for the councils voting on the contracts, the company’s own website is the primary document: it is where the question “what does this thing actually do” gets asked and answered. This piece audits that answer. Every public-facing page on flocksafety.com was read on a single day against every other page, alongside the legal policies that govern the product — and where the pages disagree with each other, both versions are quoted below in the company’s own words. Nothing here comes from a leak or an internal source, and nothing here establishes that any customer’s data was mishandled. Reporting on documents, not legal advice.

On the page Flock Safety built to answer privacy questions, under a heading that reads “Sensitive Information,” the company states its position in five words: “It captures vehicles. Not people.”

On the page Flock Safety built to sell its video cameras, the second bullet of the feature list is “People Detection Alerts.” On the page selling Flock FreeForm, the search product, the example query the company chose to advertise is “man in blue shirt and cowboy hat.” The FreeForm FAQ contains a question the company wrote itself, and titled: “Does FreeForm people search use facial recognition?”

Both pages were live on the same website on the same day. Neither links to the other.

This is not a gotcha about one stray sentence. Reading flocksafety.com end to end — the Trust center, the product pages, the industry pages, the legal policies, and the footnoted research the homepage cites — turns up a consistent pattern: the assurances are written in absolute terms, the sales copy is written in expansive terms, and the legal documents contain the carve-outs that reconcile them. A reader who visits only the Trust center comes away with a materially different understanding of the product than a reader who visits only the product pages.

Here is what the documents actually say.

The same statistic, three different sizes

The homepage leads with a number in large type: “20% of U.S. crime is solved with Flock.”

The federal facilities page closes with a stat card reading “15% US crime solved with Flock.”

The footnote on the homepage number points to Flock’s own 2025 Impact Census, and the census says something considerably narrower than either: “In jurisdictions where Flock is deployed, an estimated 20% of solved cases were assisted with Flock technology.”

Those are three different claims. “20% of U.S. crime” is a nationwide figure covering all crime. “20% of cleared cases in communities where Flock operates” excludes every unsolved crime — the large majority, by the census’s own framing — and excludes every jurisdiction without Flock cameras. The census is explicit about its method: it surveyed roughly 700 customer agencies across 43 states, anchored their self-reported estimates to FBI National Incident-Based Reporting System clearance data, and extrapolated. It is a customer survey of Flock’s own clients, disclosed as such on the page. The homepage headline drops the jurisdictional limit, drops the cleared-cases limit, and drops the survey provenance.

The daily-outcome numbers do not reconcile either. The federal page claims “2,800 Crimes solved daily.” The law enforcement page claims “2,200+ Crimes solved per week.” Those figures differ by roughly a factor of nine. Neither carries a footnote.

The missing-persons figure moves too. The homepage says “10,000+ Missing people located in 2025.” The Myths vs Facts page says Flock “found 2,000+ missing persons.” The census itself reports two extrapolation methods producing roughly 9,900 and 11,900, and a bootstrap resampling that “consistently fell between 8,000 to 19,600.” The company rounds that range to 10,000 on the homepage — defensible as a conservative midpoint, and the census says so — but the 2,000+ figure on the Myths page is not explained anywhere.

“Vehicles, not people,” and the products that search for people

The Trust center is unambiguous, repeatedly. The Data Privacy page lists what the system does not collect: “Biometric data. Driver information. Facial recognition data,” and concludes: “That means: It captures vehicles. Not people.” The Myths vs Facts page states “Flock LPR only captures vehicles, not biometric data or faces.” The What Is Flock page says cameras are “designed to capture an image focused solely on the vehicle when it passes by, deliberately avoiding the capture of unnecessary personal details about people.”

The video camera product page opens its feature list with “Get instant alerts for people and vehicles,” and names the feature: “Guardian Mode, People Detection Alerts and Visual Alerts ensure 24/7 security coverage.”

The FreeForm page goes further. Under “Search With Natural Language,” the advertised example is: “Type what you’re looking for, such as ‘man in blue shirt and cowboy hat’.” Under a heading reading “Video and LPR, All in One Place,” the copy describes reviewing “person-related video search results from enabled video feeds.” Under “Get Alerts When Your Suspect Appears,” it describes notifications for matches so teams “can assess leads without having to watch every feed continuously.” The FAQ’s own example is “woman in a black dress and cowboy hat.”

Flock’s reconciliation is a real technical distinction, and it is stated plainly on the FreeForm page: “FreeForm does not use facial recognition or biometric person recognition. Searches involving people are based on observable, non-biometric visual characteristics captured on enabled video cameras and video feeds, such as clothing and visible accessories. They are not designed to identify a specific person.”

That is a meaningful line, and it deserves to be stated at full strength: describing a person by clothing is technically and legally different from matching a face against a biometric template. Facial recognition creates a persistent identifier tied to a body. Clothing search does not.

But the distinction Flock draws in the FreeForm FAQ is between biometric and non-biometric person search. The distinction the Trust center draws is between vehicles and people — “It captures vehicles. Not people.” Those are not the same claim, and the Trust center never mentions that a person-search product exists. A reader who arrives at the Trust center with the question “does this thing watch people” is answered no. The product catalog answers yes, non-biometrically, across shared camera networks, with alerting.

“Is there a national database?” and the National LPR Network

The Trust center FAQ poses the question directly and answers it in one word:

Is there a national database of my data? No. Each agency controls its own data independently.

The products index lists, as a named product for sale: “National LPR Network — Tap into the nation’s largest crime-solving LPR network. Agencies get real-time alerts and footage from partners across the country to stop crime in motion.” The LPR camera page advertises “Join the Largest Fixed LPR Network — With billions of monthly plate reads, Flock connects communities, businesses and law enforcement in a shared network.”

Flock’s answer turns on the word automatically. Its consistent position across the Trust pages is that sharing is opt-in: “Agencies may choose to share information through a broader network, but nothing is shared automatically. Participation is opt-in, governed by defined agreements.” The law enforcement access page repeats it: “Nothing is shared unless the agency turns it on.”

Opt-in participation is a genuine architectural difference from a centrally-owned federal database, and it is the strongest thing Flock has to say here. But the question the FAQ asked itself was not “is participation automatic.” It was whether a national database exists. A queryable network spanning participating agencies nationwide, sold under the product name “National LPR Network,” is difficult to describe to a resident as “no.”

Thirty days, unless

Four Flock pages describe data retention. They describe four different retention regimes.

The Myths vs Facts page: “Flock gives customers full data ownership, auto-deleting data after 30 days.” The Data Privacy page: “Deletion is automatic.”

The LPR camera FAQ: “If a customer wishes to impose a longer retention period that is not already required by its jurisdiction’s laws, Flock will offer an extended retention for up to one year, but shall require the agency to first obtain approval from an elected official and/or a governing body.”

The What Is Flock page: data is deleted after the retention period “unless the data is required for an ongoing investigation.”

And the License Plate Reader Policy, the governing legal document, last updated June 30, 2026, contains a sentence that appears on no marketing page:

Additionally, Flock uses a fraction of LPR images (less than one percent), which are stripped of all metadata and identifying information, solely for the purpose of improving Flock Services through machine learning.

Stripping metadata is a substantive privacy measure and the policy says so. But images retained for machine-learning training are, by definition, not deleted on a 30-day rolling basis. The one-percent carve-out is disclosed in the legal policy and nowhere in the four places the company tells the public that deletion is automatic.

“Customers own their data,” and the good-faith clause

Across the Trust center and the product FAQs, ownership is stated cleanly. From the LPR page: “Ownership, title, and interest in the customer data belong to and are retained solely by our customers. Flock maintains a limited license to access the customer data for the sole purpose of providing our services.” From Myths vs Facts, on federal access: “Flock clients control their data; no secret federal agency access or ‘backdoors’ exist.”

The LPR Policy describes two categories of Flock personnel with access to customer data — “CJIS-certified engineers who are able to access CJIS data and other designated individuals who are able to access other system data (‘priveledged administrators’)” — and then sets out the disclosure standard:

Flock Safety may access, use, preserve and/or disclose the LPR data to law enforcement authorities, government officials, and/or third parties, if legally required to do so or if Flock has a good faith belief that such access, use, preservation or disclosure is reasonably necessary to comply with a legal process, enforce the agreement between Flock and the customer, or detect, prevent or otherwise address security, privacy, fraud or technical issues.

“Legally required” is a court’s judgment. “Good faith belief” is Flock’s judgment, exercised unilaterally, with no customer notification requirement stated in the policy. Both standards appear in the same sentence joined by or. This is standard commercial contract language and appears in many vendors’ terms — that is precisely the point. It is also the clause that sits underneath “customers own their data solely,” and it is not referenced on any Trust page.

The certifications that only exist as pictures

The Data Privacy page lists Flock’s compliance posture in text: “SOC 2 Type II” and “ISO 27001,” with privacy principles “aligned to ISO27701.” The LPR Policy corroborates: “Third party attestations to requirements for SOC 2 Type II and ISO 27001.”

The federal facilities page displays a badge row under the heading “Aligned to Federal Standards — Flock meets rigorous federal security and data standards.” The badges are NIST, NDAA Compliant, AICPA SOC, AICPA SOC 3, and FedRAMP.

Two of those do not appear in any written compliance statement on the site. SOC 2 and SOC 3 are different reports under the same AICPA framework — SOC 3 is a public-facing summary, SOC 2 Type II the detailed audited report — so displaying both is not itself inconsistent. FedRAMP is a distinct federal cloud authorization program with a published marketplace of authorized vendors, and it is claimed nowhere in the site’s text.

Because these are rendered as images with no alternative text in the page markup, none of these claims appear in the page’s readable text at all. A screen reader does not encounter them. A text-extraction tool does not index them. They are visible only to a person looking at the pixels — which is worth flagging for anyone reviewing vendor marketing at scale, because it is where the strongest compliance claims on the entire site are made.

Not mass surveillance, at a scale of billions

The Data Privacy page lists “Mass surveillance” and “Tracking of individuals” under Prohibited. Myths vs Facts answers “Are You Being Tracked?” with: “Flock LPR captures public vehicle info at a single point in time, not continuous movement.”

The point-in-time framing is accurate as a description of one camera. The LPR page describes the aggregate: “billions of monthly plate reads.” The homepage claims “12,000+ Communities trust Flock nationwide,” while the LPR page breaks the same footprint into “5,000+ Law Enforcement Agencies, 6,000+ Communities, 1,000+ Businesses.”

Whether a network of point-in-time captures, queried together across jurisdictions and time, constitutes something other than continuous tracking is the precise question in the Fourth Amendment litigation this journal has followed — the Norfolk case now before the Fourth Circuit, argued against the backdrop of the Supreme Court’s 2026 ruling in Chatrie v. United States, which held that Fourth Amendment protection for location data does not turn on how exhaustive the network is. The company’s marketing and its privacy assurances take opposite sides of that question on adjacent pages.

Why the brand name is the wrong unit of measurement

Everything above is about one company, because one company published it. The reason it matters beyond that company is the point this journal made in August: Flock is not the only firm selling this capability, and a rule written around Flock does not reach the others.

Motorola Solutions’ Vigilant Solutions, Genetec, Rekor, and Neology sell automated license plate reader systems to American agencies. Each publishes its own retention defaults, its own sharing architecture, and its own privacy assurances. None of them are audited by an ordinance that says “Flock.”

Denver demonstrated the mechanism in March 2026. The city’s Flock contract lapsed, all 110 cameras came down, and the coverage described a city that had removed Flock’s cameras. Denver then signed with a different vendor and continued operating an automated license plate reader network. Nothing about the surveillance changed. The brand on the pole did.

The drafting difference between a rule that survives a vendor change and one that doesn’t is documented in the statutes themselves. California’s ALPR Privacy Act — Civil Code §§ 1798.90.5 et seq. — defines its subject functionally, reaching any “automated license plate recognition system” regardless of manufacturer. That is why the pending California class action can proceed against Flock without a second statute naming Flock, and why it would reach the successor vendor if Flock were replaced tomorrow. Colorado’s failed SB 26-70 and Maryland’s HB 444 and HB 711, covered in this journal’s follow-up, were likewise drafted around the technology rather than a trademark.

The practical consequence, stated as the record shows it: a public comment, a records request, or an ordinance that names a vendor produces an answer about that vendor. One that names the capability — automated license plate readers, ALPR — produces an answer about the capability, and binds whoever is selling it next. Every question in this article is one an ALPR vendor can be asked. None of them are questions only Flock can be asked.

What this covers, and what’s still open

Everything in this piece comes from Flock Safety’s own public website and legal policies as published, read on a single day and cited page by page above. We have not obtained internal documentation, contracts, or audit logs, and nothing here establishes that any customer’s data was mishandled — a contradiction between two marketing pages is a disclosure problem, not proof of a privacy violation.

Several questions are answerable only by the company. Which retention figure governs when a customer requests the one-year extension — and how many have. How many times the “good faith belief” clause has been invoked, and whether affected customers were told. Whether the FedRAMP badge on the federal page reflects an authorization, an in-process designation, or an aspiration. Which of the two crime-solving statistics, 15% or 20%, the company considers current, and which of the two outcome rates, 2,800 daily or 2,200 weekly.

Each of those is answerable from documents the company already controls. The Fourth Circuit appeal remains pending. We’ll keep tracking both the litigation and what these pages say the next time we read them.

Sources: