This is a meaningful position from an organization I’ve previously disagreed with on ALPR transparency. The underlying disagreement was never moral — the general public should not have their data collected, catalogued, and published — but practical: the structures EFF and ACLU were endorsing produce, in the real world, the opposite of transparency.

Although both EFF and ACLU have since expressed a need and desire for stronger transparency than their previous words suggested, practical issues remain. In its new post, EFF opposes the worst version of these bills (categorical exemption) while still endorsing a framework that keeps records hidden in practice (case-by-case balancing).

The Problem

Seven states (and counting) introducing hostile legislation is not a coincidence. It’s a direct result of Flock’s lobbyists^[1] responding to public-records work that has produced policy outcomes like contract cancellations, declined renewals, and even criminal charges. EFF correctly identifies the records as “not just informational—they are leverage.”

The Washington example is instructive and worth dwelling on. A state court ruled last year that ALPR data are public records. The legislature responded by exempting them. This is the predictable endpoint of a balancing-test regime: when transparency wins on the merits in a forum that requires reasoned analysis, the response is to move the question to a forum that doesn’t.

Flock’s home state of Georgia goes further. Not content with exempting ALPR data, the state made it a misdemeanor to request or use plate data for non-law-enforcement purposes. That is the trajectory of denial-by-fee taken to its logical conclusion: when charging $5.4M for search logs (Dunwoody’s number) becomes inadequate to deter requesters, criminalization is next.

Case-by-Case Balancing and Deidentification

EFF prescribes an unworkable framework that already exists in many states: a privacy exemption requiring case-by-case balancing of transparency benefits against privacy costs.

The framework is idealistic. It is also the primary mechanism by which logs get withheld. Most states lack specific exemptions for the kind of audit data published on haveibeenflocked.com. Even states with “ALPR data” exceptions have to contend with the fact that search terms entered by a user are not “ALPR data.” Agencies often route around this by treating each search as a separate record, then assessing per-record review fees. Five- and six-figure fee estimates are common. The “balancing” and redaction in practice means an agency charges enough to make the request impossible. No balancing actually occurs.

The practical reality under these frameworks is that if you can collect enough records, you can ensure nobody ever gets to see them. That’s the opposite of the desired outcome.

The per-record framing is also a trap of agencies’ own making. If each search log is a separate record for fee purposes, it is a separate record for every other purpose too. Open records law generally requires a specific lawful basis for withholding each record, communicated to the requester, with each denial independently appealable. An agency that wants to charge per-record review fees on 10,000 “records” in a single Excel spreadsheet should be prepared to issue 10,000 individualized determinations and defend each one. Agencies want the fees, but not the obligations.

EFF’s fourth recommendation — disclosing aggregated or deidentified data while withholding personally identifiable information, and treating that process as redaction rather than record-creation — is closer to a real, and workable, solution.

Aggregation is a dodge. Counts of scans by month, hit ratios in percentages, and total-records-shared figures only tell you that surveillance is happening at scale. We know that already. Whether specific searches are lawful, whether officers are stalking exes, or whether “investigation” is being used as a pretext for anything remains locked away in a filing cabinet in the basement of the police station. Aggregate data can’t surface the Milwaukee Ayala case or the Joplin firing. Pattern-of-misuse questions require record-level data.

Deidentification is the workable part. haveibeenflocked.com already takes this approach to an extent: the site publishes audit logs but maps plates to “identifiers” to obscure their identity. This allows the patterns to remain visible — Officer X searched plate Y 124 times in two months — without exposing what plate Y is. Flock previously did the same thing with usernames in its transparency portals before stripping the IDs entirely. It’s a solution where risks like reidentification must be considered, but that’s not an insurmountable problem.

I’ve written about this approach before; haveibeenflocked.com’s “identifiers” are a partially working example of the disclosure-with-redaction structure EFF is asking legislatures to enact.

Front-loading Beats Balancing

The deeper problem with case-by-case balancing is that it does the work in the wrong place at the wrong time. Each request triggers an individualized analysis by an agency that has no incentive to perform it well, no consequences for performing it badly, and a claimed financial mechanism (per-record review or redaction fees) for converting the analysis itself into a denial.

There is a logical alternative: front-load the determination through rulemaking. My state, Iowa, has the structure largely on the books in its Fair Information Practices Act, even if implementation and enforcement are absent in practice.

Under FIPA, state agencies must promulgate rules describing what personally identifiable information they collect, why they collect it, the legal basis for collection, and which of their records are public, confidential, or mixed. The determination is made before any request arrives. The burden sits with the agency, ahead of time, rather than being shifted to the requester at the moment of request.^[2]

Front-loading also forces honest accounting. We do this elsewhere as a matter of course. If a record could contain confidential information, we treat it as if it does. Your doctor can’t store lab results in the same folder where she receives the office Christmas party invites. Yet agencies constantly argue, through public records responses, that they commingle confidential and non-confidential records and store them with third parties — Flock, email providers — that are not bound to keep those records confidential and that, in Flock’s case, will actively disseminate them to paying customers.

Strict reading of public records law makes that assumption-based structure hard. Open records statutes generally, and correctly, turn on what a record does contain, not what it could. Front-loaded rulemaking forces agencies to make the determination at the outset — at minimum, by instructing employees not to enter PII into Flock; at maximum, with a manual confidentiality justification stored with each entry.

Agencies will argue this is too burdensome; cops can’t be trusted to make these legal determinations. They will be right. The burden is the point. If an agency doesn’t know whether a search reason contains confidential information, it doesn’t know whether the search was lawful. Privacy, confidentiality, and oversight are all the same problem.

Watchlists as a Solution

Front-loading exposes one residual problem: are license plates themselves confidential PII? A categorical answer is available through watchlists.

Properly constructed, watchlist inclusion should require an active police investigation; current practice does not, which is itself part of the problem. Assuming the rule is in place, plates on a watchlist can be exempted from disclosure as part of an investigation. Plates not on a watchlist were captured and stored without an existing investigative basis. This is the “just in case” form of mass surveillance creating the biggest privacy problems.

Those historic location profiles should not be stored at all or, if they are not sensitive enough to prevent their storage, they should be subject to disclosure. The question of whether the public should see information collected without an investigatory nexus collapses into the question of whether they should be collected at all.

The watchlist approach also makes aggregation meaningful. The unit shifts from “plates scanned” — which only confirms surveillance is happening at scale — to “watchlist entries vs. open investigations,” which tells you more about how the system is being used. An agency with thousands of watchlist entries and a few dozen open investigations is using the watchlist for something other than active investigations. An agency whose watchlist entries persist for years is operating differently than one whose entries turn over in days. Either pattern is more useful for oversight than a count of plates scanned or a number of “hits.”

The contradiction is unavoidable. Either license plates are not PII and carry no privacy interest on public roads — Flock’s position when defending the cameras — or they are sensitive PII exempt from disclosure — Flock’s position when defending the audit logs from disclosure.

That contradiction is theirs, not EFF’s, and not the law’s. EFF takes the coherent position that plate data should generally be withheld from third parties, while audit logs and aggregate scan data should be public.

Flock aggressively funds the narrative that total exemption is the only solution. It would probably be right, if it didn’t assign itself the exclusive right to collect, receive, and store the very data it argues is too sensitive for the public. “The public” includes Flock.

Flock is also right that current open records laws are flawed, but the solution isn’t to hollow them out. EFF’s proposed solution is an opaque balancing test with unpredictable outcomes. My proposal is to add the requirement that governments be transparent and consistent.

Both proposals are much better than Flock’s. Neither would be required if current laws on confidentiality and open records were enforced.

False Claims and Fundamental Changes

The claim that Flock enabled backdoor access to customer data and other claims that have been in the media \— these are just false. All of the sharing that happens in the Flock system happens through permissions in the system controlled by you and your admins. It’s your data. Contractually, you control it. Flock does not own it, does not share it, does not sell your data.

@vAmy Palumbo on media claims and FOIA in February webinar

The facts Palumbo called false are well-documented on this site, as well as on Flock’s blog.

A CBP pilot ran May 9 through August 24, 2025, during the same summer in which Flock CEO Garrett Langley published a blog post titled Setting the Record Straight, where he denied federal cooperation — “not my decision, and not Flock’s decision,” he wrote — while that pilot was live.

A January 2026 Flock blog post repeated that “ICE does not have direct access to Flock cameras, systems, or data” while, in the same post, listing the CBP pilot as an arrangement that “effectively enabl[ed prospective customers] to test the product before committing.”

Throughout all of this, audit logs continued to show searches by federal agencies, while agencies like Mountain View, CA, said federal sharing was enabled without their consent.

The only time Flock accesses customer data, Palumbo went on, is to respond to “legal process” — a subpoena, a search warrant, a court order — and the customer is notified when that happens.

She made that statement around the same time Flock employees were, according to Flock, using a gymnastics room, a pool, and a daycare in Dunwoody for a sales demo. A practice Flock says it intends to continue, but “in more public locations.”

Such contradictions are routine at this point. Less than ninety seconds may be a record. Right after the “just false” line:

So where is the misinformation in the media coming from about backdoor access? It’s a misunderstanding about some of our previous pilot programs and sharing within the system. So over the summer, as these issues came to light about access from federal partners who were pilot customers, we made a lot of changes to the system in response to some of that feedback.

From “no backdoor” to a “misunderstanding” about “previous pilot programs.” A pilot program that gave access to federal agencies. Without other agencies knowing about it. If there is a falsehood anywhere, Palumbo does not mention any details.

Palumbo may dislike “backdoor” as a label, but she can’t deny the architecture it describes. Federal agencies, including immigration agencies, hooked into Flock’s sharing apparatus running searches against cameras across the country, including in states like California which restrict such use of surveillance networks.

Either way, you do not appoint a chief legal officer, expand your policy team, stand up new trust and safety programs, and create a dedicated compliance product-manager role to fix a media narrative that is false. You respond to reporters asking for comment and attach a receipt or two.

No FOIA, No Problem

Palumbo goes on a bit to encourage the, presumably, Wisconsin-based audience to apply “a balancing test” when “facing public records requests.” The information agencies are allowing Flock to collect can lead to stalking and could be misused — according to Palumbo. The dangers of releasing this information — to the public; not to Flock, its contractors, its customers, or its pilot programs, of course — can’t be overstated (when it comes to public records requests).

@vAshley Haber on product changes

The next speaker was Ashley Haber. She delivered the list of “changes to the system in response” to the supposedly false information.

Last summer, we made a change so that federal organizations at Flock are handled a bit differently than a typical state or local law enforcement agency. They are properly called out throughout the system with a label that says federal organization. They also are not included in statewide or national lookup…

Which is to say: before the summer of 2025, federal organizations were indistinguishable from local police departments in Flock’s search and sharing interface. A query run from a federal account reached every shared camera within the network’s scope, state by state, without the querying agency being explicitly flagged as federal in the audit. There was no statewide-lookup exclusion. There was no opt-out, beyond the “all or nothing” of the state- or nationwide network.

There is more.

We did a lot of work last year on cleaning up network and org audits to both be more useful for you all and also protect you all as these PRAs and FOIAs were becoming more popular and they were risking officer safety and active case investigations falling apart.

We cleaned up what we call system-generated searches so that there was not an excessive amount of records here. If someone opened a drawer, zoomed in on the map, technically on our back end it may have created another search record but it really wasn’t a user generated action… we also added the functionality starting August 8th of last year… you’re able to see if your cameras produced any results from lookup searches… the last piece is we masked certain sensitive data in response to what happened close to end of last year to protect interagency sharing.

“What happened close to end of last year” may refer to the December emails from the FBI and Flock in response to logs appearing on haveibeenflocked.com.

“Mask certain sensitive data” is the network-audit redaction feature that strips officer names, agency identifiers, plates, filters, and case numbers from the logs agencies are supposed to review for improper access.

The non-hit-search filter, added in August, lets an agency exclude queries that did not produce a plate match — and, by extension, lets them view which queries did produce a match.

So far, no agency has included this information on whether there was a hit in a network audit log. That information is not categorically exempt from open records laws, and requesters generally ask for complete logs, not ones with partial information.

“System-generated searches,” per Haber, are now filtered from the audit “to avoid confusion and too much noise,” directly contradicting Flock’s public narrative of “every search is logged.” Flock decides what is logged and what is “noise” or “confusing,” based on undisclosed and likely fluid criteria.

And finally, she discusses the “filter” that handles immigration and reproductive health searches in states that prohibit such searches.

The last bullet point on her slide informs the audience that details will remain a secret, “in order to protect the effectiveness of this feature.” Maybe Flock thinks a user of the system couldn’t figure out, without a Flock-published list, that writing ‘suspicious vehicle’ in the reason field defeats the filter.

Or maybe publishing the list would remove their ability to lie about how long they’ve been filtering any particular term. One of those explanations seems much more likely than the other.

Flock’s Solutions

Addressing false claims is easy: you correct them. Flock holds all the information and all the evidence; it has a PR team and a blog. It knows when filters went into effect and how many searches are being filtered. It’s not that hard to rebut false claims and build trust through transparency.

But if the claims happen to be true, you may have to roll out several audit-obstruction features, a bevy of new filters, and restructure your entire compliance organization.

Flock’s response is telling.

The Timeline, According to Flock

The timeline begins with an “External Claim” in March 2025, where “an individual contacted Flock with security findings after acquiring a device through illegal, unauthorized means.” Presumably, this refers to Jon Gaines’ research. A year later, Flock has not fixed those issues.

What it has done is reflected in the rest of the timeline: it “disclosed and addressed low-severity vulnerabilities,” it “responded to” the research, and it “published a response debunking false claims that the company had been hacked.”^[2] None of that fixes the issues that were disclosed to Flock in March.

The first of those actions, disclosure, happened in November, after Gaines published his report. Before November, Flock had not disclosed the issue. Not even to its customers. This is despite the requirements of the CJIS security policy, which require vendors to notify the government agency and the FBI.

The Iowa Department of Public Safety (a Flock customer and CSA for Iowa) confirmed it had received no notification from Flock. Other CSAs — the Florida Department of Law Enforcement and the Illinois State Police — did not respond to a Sunshine Act request, or asserted that vulnerability notifications are “ALPR data.”

Despite a contractual and legal obligation to provide this notification to its customers, Flock did not do so for eight months, and then only after its customers found out.

The timeline does not discuss a YouTube video before pivoting to a “second” one with “misleading claims about Flock PTZ cameras.” Flock’s timeline says it “addressed” those claims.

“Misleading Claims,” According to Flock

The “Readdressing Misleading Claims About Cybersecurity at Flock” is a lie. Not because its content is false — although it’s not exactly true — but because it doesn’t even do anything resembling addressing claims, like the section heading promises.

Can’t even trust a heading. Anyway …

I want to be crystal clear: vulnerabilities are a part of the development process of hardware and software. No company on the planet is infallible, nor is any company unhackable. It is an expected and normal process for vulnerabilities to be discovered and remediated at each stage of software development. From the point of a developer writing code all the way to that finished product running in production.

We engineer bridges and buildings so that they don’t collapse. We do all sorts of math and engineering and further science so this doesn’t happen. But occasionally, and unfortunately, they do collapse. When they collapse, we don’t shrug our shoulders and say “it’s part of the process.” We investigate the cause and address it. We make meaningful, articulable improvements to our engineering processes and standards.

We now have those collapses and their fixes codified in laws and regulations and we explain them in engineering textbooks and use them as examples on powerpoints at industry conferences and seminars. That is why we now have buildings and bridges that are more earthquake resistant than 100 years ago.

We don’t hide the problem. We don’t say it’s “an expected and normal process” for a bridge to collapse.

It’s an exceptional situation for a bridge to collapse, just as it’s an exceptional situation for a software vulnerability to be discovered in production. And just as people have died from buildings crumbling in earthquakes, people have died from insecure surveillance networks.

You don’t hide engineering issues — civil or software. You make them public, you address them, and you learn from them.

But, despite claiming that discovering these issues in production is “expected and normal,” Flock’s bulleted list of what to expect from a vendor does not include it.

The list does not mention notification or remediation for production issues. No timelines, no categories, no mentions of public vulnerability trackers, no issue categories, or anything else. Not even a “We will notify our customers and provide a remediation plan within 48 hours” or anything similar.

Flock’s Cybersecurity Team

Flock continues to invest in our team and has 10 new headcount positions slated for hiring this year, adding to our existing team of 20+ engineers. Cybersecurity is nothing without people.

This is the exact opposite of what Flock should be doing. Instead of hiring more engineers to develop more buggy AI-powered features and release more half-finished websites, Flock should be investing in hiring policy and security experts.

The post then lays out some team names without defining their headcounts, budgets, or positions in the organization hierarchy. In some companies, a 50-person “DevSecOps” team is focused on security and can shut down production when needed; in others, it’s literally one guy in Mexico City writing scripts so developers can automatically release code without review.

Castaldo does not even hint at where Flock might fall on that spectrum, and that’s cause for concern.

The “First” video

Although Castaldo omits the November video — which was Benn Jordan working with Jon Gaines — from the timeline, he devotes a section of the post to it.

In November 2025, a YouTuber released a YouTube video with two other individuals claiming to have “hacked 80,000 Flock cameras”. That statement tells you all you need to know about the credibility of the individuals and the video itself.

The video is titled “We Hacked Flock Safety Cameras in under 30 Seconds.” The closest thing to Castaldo’s quote is: “Upon further investigation, it turns out that there are over 80,000 of them. And um we got some and we hacked them.” Which is 100% true.

Blatantly misquoting an opponent’s statement before attacking it tells you all you need to know about the credibility of that individual.

Castaldo uses some choice words like “illicitly,” and “illegally” to characterize the acquisition of the Flock hardware. There is nothing “illegal” about buying hardware, and absolutely nothing suggests that Gaines (or whoever bought the hardware) did so illegally. Falsely accusing someone of criminal conduct is defamation per se in most jurisdictions.

Flock did not disclose these issues to customers. Flock did not notify customers in accordance with industry best practices and according to CJIS standards. Flock did not close out any CVEs, nor did it open any new ones. Flock did not tell Jon Gaines “we are aware of this and we will fix it.” And at no time in 2025 (or at all, for that matter) did Flock communicate a fix.

While the findings were legitimate, they were all of low severity. Meaning the risk to customers or customer data was near zero.

Of the findings in the report, many are high severity when going by the framework laid out by the U.S. Department of Justice, which governs much of the data. Castaldo does not specify what framework he uses for his “low severity” classification or his “near zero” risk assessment.

Dunwoody gave us vibes-based auditing and compliance, Castaldo adds another layer: vibe-based cybersecurity.

Had this individual not prevented [the camera] from connecting to our cloud, most of their findings would have been moot.

This is a fair enough statement in isolation, but does not address the two key problems.

First, there is no evidence that Flock discovered and fixed these issues, and rolled out an update. No required customer notifications, no proactive security disclosures, nothing. Complete silence.

If these issues were indeed fixed, and were not the result of plain negligence, nothing is lost by publishing these issues. Most software vendors do exactly that to build trust. Microsoft, for example, has a page called “Vulnerabilities and Exploits” on its main website, and it includes a list of fixes with each update, including any security fixes.

If Flock had published anything or notified anyone, cross-referencing those notifications against Jon Gaines’ report would make for an easy exercise in ticking off fixed issues and seeing what — if anything — remains.

Flock could easily restore trust and show that it is on top of its security by publishing a few emails that it already sent to its customers when it first discovered these issues — as it is required to do — or when it fixed the issues — as is standard practice.

Second, there have been no patches for this particular operating system since 2021. While security issues could have been deployed for Flock’s custom software, no vendor OS fixes were released.

Connecting it to the network would not have caused non-existent patches to be applied.

The “Second” video

This individual did not ethically submit any information to Flock prior to the release of their video

If I’m recalling the video correctly, it is true Jordan did not submit information to Flock prior to the release of the video. The last time issues were disclosed to Flock — in March, according to the timeline — they were not fixed or disclosed even months later (or, to this day, as far as I’m aware). Disclosure to a vendor is often the right choice, but there are no bright lines in ethics.

In this case, anyone whose ethics dictate minimization of harm would have done exactly what Jordan did. He denied Flock a second opportunity to jeopardize people’s safety by trying to bury an issue, as they did when issues were disclosed to them in March.

Just Keep Digging

Flock worked with our carrier partner to quickly resolve the network configuration issue. … Flock has also modified the diagnostic interface to require our technicians to log in with a username and password. Again, this interface is intended to be usable when a technician is physically present.

First, let’s address that the software had to be “modified” to require a username and password.

According to Castaldo’s post, Flock did all of these things:

• “Threat modeling during the design phase of a product”
• “Scanning and fixing code as the developer is writing it”
• “Scanning and fixing finished code when a developer submits it to the code repository”
• “Scanning and fixing applications running in production”
• “Continuously scanning and monitoring the infrastructure the application is running in”
• “Conducting penetration tests against all of the above.”

To top it off, he writes immediately below that list: “There is a cliche about cybersecurity being an onion with many layers, and that remains accurate today.”

Yet, in that whole development process, nobody at Flock, at any time, said: “hey, maybe we should require a username and password.” Even hardcoding “DonkeyKeepOut!” as a password would have prevented Jordan from gaining access.

The second issue is that no matter what layers Flock might have in its development process, there was only one in its security: Verizon’s configuration. In this, Flock’s security model is more like banana: a single layer that can easily be peeled away by anyone who wants access.

Flock gave Verizon the unchecked, unreviewed, unsupervised, ability to create and manage the security configuration for an interface that was not secured with a password.

Even without a “misconfiguration,” Verizon employees would have had access. A company with roughly as many employees as Burbank, CA has residents (plus who knows how many contractors) having unfettered access to live videos of kids playing in parks is Castaldo’s baseline definition of secure.

On Android

The software on Flock’s cameras hasn’t received vendor security updates since 2021. That is the central fact of this section of Castaldo’s post, and the one he does not address. Instead, he offers several paragraphs of technically misleading context about chip architectures — context that, on examination, actually makes his position worse.

Flock hardware runs on a heavily modified version of the Android operating system maintained by Google. This is an open-source operating system, meaning anyone in the world can look at the code and use it.

Flock has “heavily modified” Android, but never published those modifications. Yet we should feel assured — presumably based on vibes — that its “heavy modifications” are not material enough to affect security.

This is very different from the CPU in a computer running Windows or MacOS. Qualcomm’s chipsets are purpose-built and support specific operating system versions.

This is somewhat backwards, because hardware vendors don’t tend to build chips to accommodate operating systems, but it’s accurate enough in the way it matters: there is a fixed relationship between the hardware and the OS.

Flock Falcons reportedly use Qualcomm Snapdragon 625 chips, which are early 64-bit ARM chips (like the M1/M2 chips in current Macs). These were supported by Android until version 8.0 or 8.1, support for which ended in 2021. This is the same as support for older Intel-based Macbooks, which is also ending. There is nothing particularly unique or different about Qualcomm chips in that regard.

It’s theoretically possible that for the past five years, Flock has been paying engineers to backport security fixes to this unsupported version of Android. There are projects like LineageOS that do exactly this to support aging phones in primarily low-income countries.

It’s also theoretically possible that Flock designed the Falcon around 2017 around the then-popular Snapdragon 625, and that it did not replace all of its devices in 2021 when supported ended, but instead designed an entirely new line of devices (which it called “Flock LPR”), with the goal of replacing the Snapdragon 625-based Falcons as they age out of service.

Qualcomm produces a custom, heavily modified version of Google Android that is designed to run on their chipsets.

Qualcomm does produce a modified Android that is optimized for its hardware, this much is true. The problem is that Qualcomm takes an official Google Android version and modifies it for its hardware.

Qualcomm released its last full BSP for the Snapdragon 625 in 2019, and its last security update in Q4 of 2020.

Android Things

Gaines’ security report finds a problem in “Android Things 8.1” being EOL. Android Things was a popular OS for the Snapdragon 625. In the blog post, Castaldo emphatically bolds that “Flock has never used Android Things, in any product.”

Never mind that it contradicts the earlier “all of the findings were previously discovered by Flock’s cybersecurity team,” or that this is the first time Flock has raised the point, the distinction between “Android Things 8.1” or “Android 8.1” is irrelevant.

Because “Qualcomm’s chipsets are purpose-built and support specific operating system versions,”^[3] none of those “specific operating system versions” have been supported since 2021. Not Android 8.1, not Qualcomm’s BSPs, not Android Things 8.1.

Even if the statement were true — which I doubt, because I trust Gaines and Jordan to be able to identify an OS — it would be a nice “gotcha” on an entirely meaningless fact.

At the end of the day, the software hasn’t received security updates since 2021. That’s the point that matters, and the one Castaldo does not address.

Backporting

We will continue to backport any necessary security patches, as required under our agreements with all customers.

If Flock is indeed backporting security patches to Android (Things) 8.0 or 8.1, or whatever the case may be, then security itself may not be the issue. However, “as required under our agreements with all customers” includes the requirement to notify customers when they do discover security vulnerabilities.

Each time Flock backports a fix, its contracts — at least those with CJIS security addenda, which should be all government contracts — require notifications to be sent to contracting agencies (and the FBI). No notifications have ever been sent out.

The other problem is that Qualcomm’s proprietary modifications to Android, which Flock just explained are tied to the hardware, are not open source at all. There is no backporting fixes to those parts of the OS.

Third party attestation

Yes, Flock has qualified third-party attestations of its cybersecurity. What you should also expect from your vendors is continuous audits by qualified, third-party firms. Flock takes this seriously and goes far beyond surface-level audits.

The post rattles off a list of security standards or frameworks, this time omitting HECVAT and FERPA, and points to its “trust center” where, “[o]nce you gain authorization for access, you may review” the relevant documents.

But you don’t need access to see that the list of actual certifications — SOC2 Type II, ISO 27001, ISO 27017, etc. — are about organizational and procedural controls, not software vulnerabilities.

Flock “maintains standards” of “CJIS Insights”, “CJIS ACE”, “FedRAMP 20x,” and “NDAA”. “CJIS Insight” (singular — Flock can’t even get the product name right) is a compliance-tracking software dashboard sold by Diverse Computing, a company in Tallahassee, Florida. “CJIS ACE” is a commercial compliance assessment also sold by Diverse Computing. Neither is a government certification, and neither is affiliated with the DOJ or the FBI.

This is where it gets really interesting and where we have to break out our diamond pickaxes.

Castaldo spent most of this post assuring us that their use of an outdated operating system is fine because they backport software. Now he invokes CJIS and NDAA.

CJIS requires the use of FIPS-140 validated encryption modules. FedRAMP — which Flock also claims and which was codified into law by the NDAA — independently requires FIPS-140 validation as well. To the extent Flock has FIPS-140-2 validation, it has never produced documentation to my knowledge. Soon — in September 2026 — FIPS 140-2 will be no more. Flock will need to move to FIPS-140-3.

FIPS-140-3 places stricter standards on the “Operational Environment,” which includes the operating system: Flock will have to validate the combination of obsolete hardware (Snapdragon 625) and custom operating system as a single “hybrid module.” So far, such a hybrid module does not show up in NIST’s database.

As previously reported, Castaldo’s co-founder at “Security Tinkerers,” Will Lin, sits on the board of Bishop Fox — the firm Flock hired for its security audit. Castaldo mentions Bishop Fox only once in passing in this post, and does not mention this relationship at all in the section about third-party verification.

The Proof

I have called for this before, and I will call for it again: Flock should publish its actual NIST validation certificates, and its security disclosures to its customers.

Castaldo’s 2,000-word defense does not contain a single customer notification, a single CVE, or a single NIST certificate number. It relies on strawmen arguments, mischaracterizations of hardware lifecycles, and a little light defamation.

Stop digging and start fixing.

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.

Where Hotlists Come From

First, the official watchlist: this comes from the FBI’s National Crime Information Center (NCIC), which is “a computerized index of missing persons and criminal information and is designed for the rapid exchange of information between criminal justice agencies”.^[1] The FBI’s database was created in 1967, and has since grown tremendously in scope.

NCIC stores information about property (vehicles, firearms, etc.) as well as people (e.g., warrants, gangs, terrorism, anyone of interest to the secret service). This information can then be queried by state and local agencies, in accordance with the FBI’s terms, which are laid out in federal regulations and the CJIS Security Policy.

The information from NCIC — which is often inaccurate — is uploaded by states, after which the federal government and its state partners make it accessible to federal and local agencies via portals like Palantir’s ICM and Falcon, and, to the point of this article, Flock’s FlockOS and its “hotlists.”

Until recently,^[2] Flock had a blog post on its website called “What Happens When a Wanted Car Passes a Flock Safety Camera?” That post appears to have been deleted in a hurry. It is still linked to from Flock’s other blog articles, including one that describes the “Hot List feature.”

Because Flock has requested that sites like archive.org’s Wayback Machine not archive the content on its site, an archived copy is unavailable, but here is a paragraph from a referencing post:

Flock Safety’s Hot List feature allows businesses to receive alerts on stolen vehicles, known wanted criminals and more, as long as the private business chooses to share their LPR cameras with law enforcement. Law enforcement will then be automatically notified if a vehicle associated with the NCIC enters your property, allowing them to take action. It works via an integration into the Federal Bureau of Investigation’s National Crime Information Center (NCIC), which is updated daily.

Customers can also place at-risk vehicles or license plates on custom Hot Lists to alert law enforcement or on-site security instantly if/when they enter their property. — “Vehicle and Catalytic Converter Theft: Flock Safety’s Solutions for Businesses,” Flock Blog, May 14, 2025

The “Customers” referred to are Flock’s commercial customers.

The Unregulated Layer

This is the second category of watchlist: managed by local police agencies, HOAs, neighborhood associations, and businesses, and shareable without constraints — but somehow still exempt from public records law, according to police agencies.

NCIC entries, which have specific retention policies and restrictions on sharing, are often copied to custom log entries in the Flock system.

Other entries, like the one shown on the “SUS Vehicles” hotlist, can be for vague reasons like “Poss vin swap, stop and verify” — suggesting the vehicle has been placed on the list to be stopped without probable cause, or, at best, for pretextual reasons.

These stops are then combined with departmental policies to keep their use secret, like those seen in Wapello County, Iowa:

DO NOT MENTION ALPR USAGE TO THE OCCUPANTS OF THE VEHICLE. Simply reference that you ran the plate and observed an NCIC hit … DO NOT MENTION ALPR USAGE IN YOUR REPORT OR COMPLAINT UNLESS ABSOLUTELY NECESSARY. … If it is necessary to explain in a report, it is advised to use language such as “Using County resources, I discovered the suspect vehicle was bearing Iowa plate…” — Wapello County Sheriff’s Office SOP III-20, November 7, 2025.

Whether a stop was conducted based on a watchlist hit can only be established if a defense attorney asks the question and the police answer it. That may not happen until the victim of an unlawful stop has spent days, weeks, or even months in jail.

What Counts as a “Reason”

I examined 800 “Custom Hotlist” entries entered by Harris County Constable Precinct 5^[3], which covers over 300 square miles of western Houston suburbs — an area with 1.3 million residents, policed by roughly 450 sworn deputies. 101 of those deputies — nearly one in four — appear as users in the hotlist logs. 51% of entries had no listed case number and 73% were set to never expire. The criteria used varied from outstanding warrants (copied over from NCIC), vague suspicion that a VIN may be altered, or no reason at all.

An investigator at the Precinct 5 constable’s office used Flock to locate and seize a vehicle for an alleged theft that another county refused to investigate, that was never entered into any criminal database, and whose statute of limitations expired seven years earlier. The hotlist entry reads:

Vehicle is not listed as stolen in NCIC/TCIC. Registered owner claims vehicle was stolen out of Colorado County in 2012. Colorado County refused to enter vehicle as stolen. Find PC and stop/identify all occupants… Vehicle will be seized for a Civil Seizure hearing. Do not release vehicle

Investigator James Dancer created this entry on July 3, 2024, for a Texas plate ending in –981^[4], case number 2407-00085, with a one-month expiry of August 2, 2024.

The vehicle was allegedly stolen twelve years ago, according to a claim by the registered owner. Another agency — Colorado County — had already declined to enter it as stolen; possibly because the statute of limitations expired seven years earlier, in 2017.^[5]

Colorado County already decided there was no criminal case here. Dancer’s entry instructs his peers to manufacture a justification for a stop (“Find PC”) and seize the vehicle anyway. That outcome (“Do not release vehicle”) is predetermined in the hotlist entry. Once the vehicle is seized, civil forfeiture requires only a “preponderance of the evidence” that the vehicle is connected to criminal activity.

The vehicle, which is at least twelve years old, won’t be returned to the registered owner. Instead, it will be sold at auction and the proceeds will go to Dancer’s department and the Harris County district attorney. The Institute for Justice has previously filed a class-action suit for exactly this type of abuse of the civil asset forfeiture process in Houston.

Dancer’s entry also contains an instruction that appears 47 times across the 800 entries: “Find PC and stop/identify all occupants.” In 31 of those, the instruction was created by Kayla Cohan (formerly Fesperman) using a near-identical template: “BMV Susp Vehicle- BLK FORD BRONCO-Develop PC, Stop and ID Occupants.” Burglary of a Motor Vehicle (BMV) is typically a misdemeanor. These entries instruct officers to first locate a vehicle through Flock’s automated surveillance, then find or develop a legal pretext for the stop afterward — inverting the Fourth Amendment’s requirement that probable cause precede the seizure.

There is no reason to believe that the “occupants” of these vehicles, who are entirely unknown at the time the entry is created, are involved in the crime the driver is suspected of committing. That crime is equally unknown until someone “finds PC.”

Similarly, another one of Dancer’s entries for a vehicle that could be related to shoplifting at Bath and Body Works contains a slightly more detailed instruction:

Bath and Body Works Shoplifting Suspect Vehicle/ Find PC, Stop and Identify all Occupants. if any arrests are made, contact Investigator James Dancer (5I10), will most likely make scene. Contact Pct. 5 Dispatch (281) 463-6666. Send all questions/ information to James.Dancer@cn5.hctx.net

The investigator email addresses and direct phone numbers embedded in these reason fields serve a secondary purpose: when hotlists are shared between agencies, the receiving agency also receives internal contact information for the creating agency’s investigators — an informal contact-sharing network piggybacking on the surveillance system.

Another vehicle is put under indefinite surveillance by Andrea Trevino. No case number given: “SURVEILLANCE ONLY//DO NOT MAKE CONTACT.”

Jose Ramos added 6 vehicles for “Tolls” — five with no case numbers, one with case number “N/A”, all set to never expire.

Mental health crises can land you on the watchlist: “SUICIDAL GUY” (one month), “WARRANT AND SUICIDAL GUY” (no expiration), “SUICIDAL ARMED” (one month), “Welfare Check” (one month). None had case numbers.

You can also earn a permanent watchlist spot for how your car sounds: “Engine does not sound stock;” or for cryptic reasons like “300 has a badge”, “Memorial Mall”, or simply because the sergeant said so (“Sgt request”, case number: “theft warrant”).

Other entries include “SUS”, “sus veh”, “susp”, and “fug”.

A dashboard from an Iowa agency shows that even where reasons are entered, they indicate problems; there is a hotlist for “Sex Offender” — a conviction is not justification for indefinite surveillance — as well as a hotlist for “Protection Order.”

That latter category’s existence is surprising, given that Arkansas and Wisconsin police failed to act when a person with an ankle-monitor passed at least one Flock camera, as he went to the home of a 16-year-old who had a protective order in place. The pair then disappeared for two months, until someone spotted them at a truck stop in Nebraska.

Sharing Without Scrutiny

When hotlists are shared, the receiving agency receives an automated email from Flock. That email considers the data sharing to be “Great News!” and provides instructions on how to start using the shared hotlist.

What it does not provide is information about what is on the list — including whether the hotlist contains any information sourced from NCIC, which would be restricted by law. It also does not provide information about the policies in place for maintaining and populating the list.

Flock sent the email above to a Minnesota agency. The email does not mention that Minnesota law^[6] says ALPR data “may only be matched with data in the Minnesota license plate data file” or when related to an active criminal investigation. Instead, Flock cheerfully directs its agency customer to accept the mystery hotlist from Florida — in a way that more likely than not violates Minnesota law — and thanks them “for being part of the Flock.”

Although Transparency Portals include a “Hotlists Alerted On” column, Transparency Portals don’t appear to show custom hotlists.^[7] Flock’s transparency portal configuration tool, however, shows that “This value will be automatically generated according to your Flock settings.”

Flock Won’t Say How Often It’s Wrong

A common concern surrounding the use of automated enforcement is accuracy — we know that misreads happen, but we don’t know how often. And Flock won’t say.

Flock monitors and collects misread rates; it declined to provide Business Insider with specifics about the data. When customers flag misreads, that data is pulled into the company’s training set to improve its model, and the company works with local law enforcement to understand the cause of the incident, a spokesperson said. — ‘Flock Flocked up’, Business Insider, March 9, 2026.

What Flock has said, in its undated^[8] blog post “Assessing the Accuracy of Computer Vision Methods for Traffic Data Collection,” is that “Flock Safety cameras correctly identified 92.3% of vehicles by classifying them across six vehicle categories via computer vision, in accordance with the FHWA’s 13-bin classification system.”

The FHWA 13-bin system uses broad categories like “Motorcycle” or “Five-axle single-trailer trucks.” It has several categories for tractor-trailer combinations with differing numbers of axles. Flock doesn’t specify, but most likely reduced the number of bins to six by collapsing several vehicle classes into a single “truck” classification.

That 92.3% figure measures vehicle type classification — whether the system can tell a motorcycle from a tractor-trailer — not plate-reading accuracy. If Flock’s cameras fail to correctly classify vehicle type in nearly 8% of cases using these broad categories, their accuracy at reading individual plate characters or matching the more granular attributes in its FreeForm search — clothing, dents, and other vehicle and person characteristics — is anyone’s guess. Flock’s refusal to disclose plate-read accuracy or submit to independent auditing leaves that question unanswered.

Even NCIC Gets It Wrong 77% of the Time

But even when accuracy is discounted, an overview of Axon data from the Story County, Iowa sheriff’s office^[9], shows that matches are often a bust. In Story County’s “Erroneous hotlist hits” report, which covers approximately a month, hits were only sourced from NCIC.

In that month, the sheriff’s office reported 214 incorrect hits. 165 (77%) were flagged “Wrong state,” 10 were “correct” but “No action taken,” “No Associated Party in Vehicle” or “Parked - Unoccupied,” 7 were “Dismissed,” and 3 were “incorrect.” The report does not include information about the total number of scans, or the number of accurate hits.

77% of vehicles from NCIC being matched to vehicles from the wrong state is an issue. The issue is further compounded by NCIC not distinguishing between characters like “O” and “0” or “I” and “1”. And that’s NCIC, which has policies on accuracy and review of information.^[10] For Flock’s “Custom Hot Lists,” all it takes for someone in Minnesota to get pulled over is for Brittany Smith from Florida to enter an incorrect digit.

Even if the technology were 100% accurate, which it isn’t, it is still subject to the principle of “Garbage in, garbage out.”

Who Else Gets the Data

After an entry is created or copied into a custom hotlist, users can opt to receive notifications through Flock’s mobile app. If the entry is criminal justice information (CJI) copied from NCIC, it may only be accessed through a secure workstation.

Instead, Flock encourages use of its mobile app, which offers simple on/off toggles for receiving alerts. It also offers notifications based on a user’s location, suggesting Flock mobile app users’ locations are being recorded and tracked.

The risks of inaccurate or poorly-maintained hotlists to the public are well-documented. In Sherwood, Arkansas, a Flock ALPR misread a plate obscured by a loose plate holder. Police ordered a couple out of their car at gunpoint and handcuffed the woman in front of her children. In Morristown, Tennessee, a Flock camera misread an “O” as a “0,” and the Herron family — with their 3-year-old granddaughter in the car — was pulled over at gunpoint. In Toledo, Ohio, a misread “7” became a “2,” and Brandon Upchurch was mauled by a police dog and jailed. And in Aurora Colorado, an unidentified system flagged an SUV as a stolen motorcycle from another state — the plate number was the same, but it was the wrong state, like in 77% of Story County, Iowa’s erroneous hits.

But members of the general public are not the only ones at risk — police officers are too.

The hotlist data itself is federally regulated under the CJIS Security Policy:

[W]hether it’s bring your own device (BYOD) or a agency-issued phone, [law enforcement] must follow the protocols set out by the FBI’s Criminal Justice Information Services (CJIS) compliance for mobile device security and adhere to the CJIS MFA requirements that go into effect October 1, 2024 — Essential guide to agency-issued phones and BYOD policies, Verizon Business, September 13, 2024.

But those rules don’t extend to the data Flock collects through the app — like location data. It can collect or resell that data, or use it in its other products. There is no requirement for Flock to safeguard that data, or to screen the people — employees and contractors — with access to that data.

How much of the data Flock transmits is unclear, but the mobile app has a direct integration with FullStory^[11], a “Behavioral Data & Digital Analytics Platform.” It is equally unclear where the data goes from there.

We have already seen the predictable outcome when Flock failed to secure an API key and leaked live location data, including this exact “Officer mobile app location data (phone, smartwatch).”

With no functional safeguards in place, a history of leaks, and no consequences imposed by the Department of Justice or state agencies, it is left for individual officers to decide if they want Flock to enter their realtime locations — potentially both on-duty and off-duty — into its database, and whether all of Flock’s employees, contractors, and partners should have access to that data.

No Recourse, No Oversight, No End Date

There is no mechanism to discover which agencies are using what custom hotlists, or for a person to discover whether they are on a custom hotlist, to challenge their inclusion, or to request removal. There is no judicial review, no expiration requirement, and no independent audit.

The entries examined here — 73% set to never expire, half with no case number, many with reasons no more specific than “SUS” — are not aberrations in an otherwise functional system. They are the system functioning as designed.

Flock’s marketing promises transparency and accountability. Its logs consistently tell a different story: indefinite surveillance authorized by a single officer’s keystroke, shared across jurisdictions without review, immune from public records requests, and enforced through pretextual stops that its own users are instructed to manufacture.

The question is not whether the technology works. The question is whether anyone is watching the people who use it.

The answer, to date, is “No.”

State and local governments aggressively resist open records requests related to Flock. They will apply any exemption, no matter how non-sensical. Part of that is simply the government’s mindset— the less accountability to the public the better. Part of that is Flock.

The Guide

Flock issues informal guidance to its customers on how to handle open records requests. A document, “Guide to Flock Safety Data for Open Records Law” (last updated September 2025) opens by telling the reader that they do not have to create records.

@Guide to Flock Safety Data for Open Records Laws

The guide offers wildly incorrect legal advice.

Generally limited disclosure [for ALPR data] across most states. Some states exempt all data captured by or derived from any automatic license plate reader system from disclosure either by express statute or per case law

Rather than there being “generally limited disclosure”, few states have express protections for ALPR data. This is self-evident from Flock’s wholly unregulated status as a provider of “photos taken on public roadways where there is no expectation of privacy.” Those same photos are not “generally” exempt from open records requests.

Agencies should consider whether to redact license plates, search reasons, and case numbers from these logs, as well as other potential fields that may be deemed sensitive

Similarly, the open records laws that I am aware of do not permit redaction of “fields that may be deemed sensitive.” Rather than relying on a clerk or a cop to subjectively deem something to meet an unspecified standard of sensitivity, open records law tends to only permit redaction of items that meet specific objective criteria defined by statute.

Flock’s document goes through every category of public record related to its system to identify possible exemptions, suggesting ones for police investigations, security exceptions, and privacy reasons. Throw exemptions at the requester and see what sticks.

When all else fails, Flock suggests in its suppression manual dressed up as customer support, its government customers should not disclose the record, like the law requires, but “consider negotiating a narrowed timeframe” and charging fees.

We’ve since seen other guidance, where customers are instructed to keep searches “as vague as permissible” come from FBI agents in Flock’s hometown of Atlanta.

In recent months, there has been a marked uptick in audit logs submitted to haveibeenflocked.com in garbled PDFs—including from states that require public records to be produced in their original electronic format.

One agency in Arizona did the FOIA-thing and printed out the CSVs to scan them right back in. Another delivered them to the requester on paper. Perhaps Flock updated its guidance.

Contractual Obstruction

Flock does not restrict itself to advising customers on their open records process, it inserts itself. Some, but not all, contracts create a duty for customers to notify Flock and delay open records responses.

If a request is made pursuant to the Iowa Open Records Act, Iowa Code chapter 22, to examine Confidential Information identified herein, the Customer will notify Flock. Flock will be given not less than ten (10) calendar days within which to file an action in the Iowa District Court… seeking the entry of a declaratory order or injunction to protect and keep confidential the information identified as confidential herein. — Johnston, IA Service Agreement

This is clearly problematic from a transparency perspective, and, raises serious legal questions in Iowa. The Iowa Open Records Act does not set fixed timelines for responses—it requires governments to respond "promptly."A mandatory minimum ten-day response delay is not “promptly.”

In Grafton, WI the language is a bit softer; there, Flock requires “reasonable prior notice.”

That’s not the only problem with this clause though. It is the vendor assuming a decision-making role in the non-delegable public records process. While a government may delegate some of its functions to third parties, the duty to respond to open records requests “promptly” lies exclusively with the custodian, and delegating such decision-making authority to a private party is likely unconstitutional.^[1]

It’s the same reason we can’t hire mercenaries to police our cities—government employees must be accountable to the people. At least on paper.

Nothing the Public Can Gain

Then there are the “transparency” portals. Flock has stripped functionality to hide essential information, but cities often still refer people who request access to log files to the portals. In public, Flock sells it as a transparency tool “to promote trust, accountability, and citizen privacy in policing.” In private, Flock tells its customers the truth: it’s useless transparency theater.

Take a look at this sample Transparency Portal and let me know if you’d like anything changed. All fields can be edited, deleted or added to. Any of the fields in grey indicate information that will be pulled directly from your Flock account. The only other thing worth noting is the Search Audit…I have attached an example. There is nothing the public can gain from this report, as it only provides the search date, camera and search reason. — Email from Libby Landers, Flock, Senior Customer Success Manager, to Ridgecrest, CA Police Chief Ysit (June 25, 2024)

@“Nothing the public can gain” - Ridgecrest, CA (2024)

If you are inclined toward charitable interpretation, you could see this as an unfortunately-worded email hastily typed by a customer service rep with an inflationary “Senior” title. Fair.

Except the same email Libby Landers sent to a California police department in 2024 shows up nine months later, word for word, in Prosser, WA, with someone else’s signature (Danica Pierce, Flock’s Local Customer Success Manager I).

@“Nothing the public can gain” - Prosser, WA (2025)

Someone at Flock approved the message for use as a form email.

The next sentence in the form email is also worth highlighting:

However, if you find your department’s users are not consistently searching off of incident/case numbers, that may be a reason to hide the Search Audit. It is entirely up to you but just like to point this out.

It relates to another email, where Flock cites the CJIS Security Policy:

Per legal: A case number and/or call for service number listed for the search reason is a Flock Requirement + Best Practice and required under Criminal Information Services (CJIS) Security Policy as promulgated by the FBI.

4.2.5.1 Justification In addition to the use of purpose codes and togging information, all users shall provide a reason for all all inquiries whenever requested by NCIC System Managers, CAs, local agency administrators, or their representative

— Email from Kyle Turner, Senior Customer Success Manager, Flock to Ridgecrest, CA Police Chief Groves (Feb 2025) (emphasis in original)

@Email Flock to Ridgecrest, CA Police Chief Groves (Feb 2025)

In a form email, Flock tells its customers to hide the evidence if its customers plan to violate their contracts with the US Department of Justice and federal rules and regulations (and, in many cases, parallel state law).

Ridgecrest, CA PD has disabled case numbers in the Transparency Portal.

The Lawyers Know

The government, or, at least, its lawyers, know that their legal justifications for denying requests are thin. In an email exchange between Prosser, WA, city officials and (presumably) their attorney, sparked by a records request from MuckRock user Rose Terse, the attorney expresses some frustration with Flock’s relationship to public records.

Emily Guildner of Thompson, Guildner & Associates, writes to her partner, Nikki Thompson:

i think it is a better argument that its not a record yet but i really just want all of our clients to stop using flock cameras.

She attempts to come up with a justification but comes up short:

I guess the question is whether it is “a writing” already out there but in an illegible format or not. Cities do have to pull data from a data base if requested, I just don’t know what form this is in, or if its in no form until there is an inquiry run?

She settles on the poorly-fleshed out theory regardless:

well our position on these is that they are a little different in that the pictures etc are records that are out there, the audit logs are not a record yet. so its not about access its about the fact that we would have to create a record to respond to the records request. but i think we’re on thin ice…

Thompson finally sends a proposed response to city staff, denying the request for it being “creating a record” and asking staff:

Thoughts? Are you sure you don’t want to turn [the Flock cameras] off? Remember that attorney fees are mandatory, if a City loses in public records litigation.

Two months later, in January 2026, Prosser turned them off.

@Prosser, WA attorney email chain

Flock industrialized existing government hostility to public records with guides, form emails, contract clauses, and a “transparency portal” engineered to disclose nothing of value.

Prosser’s attorney asked the right question. More cities should answer.

Cross-posted from Footnote 4A, where I cover Flock, privacy, and public-private surveillance infrastructure more broadly. Flock-specific posts live on haveibeenflocked.com.

You can request the full, current list from any agency using Flock—federal policy requires them to have it available; the company’s concern about speculative “officer safety” scenarios apparently does not extend to its own employees, whose names and signatures are being filed into public records as a matter of course — hopefully with their knowledge and consent.

To be clear, the “illegible” signatures are completely illegible—but apparently still sufficient for Flock, Story County, Iowa, the Iowa Department of Public Safety, and the FBI.

These 28 Flock employees signed the following statement:

I hereby certify that I am familiar with the contents of (1) the Security Addendum, including its legal authority and purpose; (2) the NCIC Operating Manual; (3) the CJIS Security Policy; and (4) Title 28, Code of Federal Regulations, Part 20, and agree to be bound by their provisions.

I recognize that criminal history record information and related data, by its very nature, is sensitive and has potential for great harm if misused.

I acknowledge that access to criminal history record information and related data is therefore limited to the purpose(s) for which a government agency has entered into the contract incorporating this Security Addendum.

I understand that misuse of the system by, among other things: accessing it without authorization; accessing it by exceeding authorization; accessing it for an improper purpose; using, disseminating or re-disseminating information received as a result of this contract for a purpose other than that envisioned by the contract, may subject me to administrative and criminal penalties.

I understand that accessing the system for an appropriate purpose and then using, disseminating or re-disseminating the information received for another purpose other than execution of the contract also constitutes misuse.

I further understand that the occurrence of misuse does not depend upon whether or not I receive additional compensation for such authorized activity. Such exposure for misuse includes, but is not limited to, suspension or loss of employment and prosecution for state and federal crimes.

This certification, along with a fingerprint-based background check, is a requirement under the CJIS Security Policy:

This section’s security terms and requirements apply to all personnel who have unescorted access to unencrypted CJI. Regardless of the implementation model – physical data center, virtual cloud solution, or a hybrid model – unescorted access to unencrypted CJI must be determined by the agency taking into consideration if those individuals have unescorted logical or physical access to any information system resulting in the ability, right, or privilege to view, modify, or make use of unencrypted CJI. — CJIS Security Policy, v5.9.5, § 5.12, p. 212.

Note that the policy is explicit about “logical or physical access.”

Iowa DPS puts it in even clearer terms in a guidance document:

All private contractors who perform criminal justice functions shall acknowledge, via signing of the Security Addendum Certification page, and abide by all aspects of the CJIS Security Addendum — Iowa DPS, Requirements Document FBI CJIS Security Policy Version 5.3",^[1] p. 9

There only being 28 employees who would need to certify is … at best, implausible.

Yet, after months of back and forth between Story County and the Iowa Department of Public Safety, this is the list Flock and the county attorney produced.

Who is Missing

Any Flock employees with access and a first name that starts with D–Z. Unless there aren’t any, but that seems improbable.

We know that Flock “LPR” cameras contain unencrypted photos and videos. The CJIS Security Policy is clear that anyone with physical access to CJI should be on the list; that would include all (subcontractor) installers. The alternative, that the footage stored on the devices is not CJI, renders it non-confidential and, in most states, a public record.

For the reported issue where Flock cameras were publicly exposed on the Internet, Flock’s Chief Legal Officer Dan Haley has downplayed the severity of the “not a hack” by claiming it was a “configuration error” perpetrated by Verizon.

If Verizon’s employees can configure the system to expose the information, they have access sufficient to trigger the certification requirement. Flock does not consider this to be a security incident, implying Verizon personnel have authorized access, yet they do not appear to be on the list.

The list should also include of all Flock’s Upwork contractors, whoever has access to its Danish screen-recorder, and, assuming these are Flock’s own accounts, anyone using Flock City PD - Law Enforcement Sales, Flock City PD - Law Enforcement Sales Demo, Flock RTCC, Flock Safety Admins, Flock Safety Customer, Flock Safety Engineering, Flock Safety Sales, Flock Safety Campus Security Training, Flock Safety LE Training, Flock Safety Sales, and Florida LE Flock Training, which all run on production data (i.e. real people’s movements are regularly being searched for Flock’s sales and training purposes).

Notably, Robert Otten, Flock’s “Head of Security, Risk and Compliance” (or similar titles), attested to each of the 28 signatures but did not certify his own adherence to the CJIS Security Policy. A suspicious absence, if the list were complete.

What is Missing

Around 6,000 contracts, based on Flock’s reported number of government customers. These certifications are tied to specific CJIS addenda, which are tied to specific contracts, via “the contract incorporating this Security Addendum.” Each person on the list needs to read each of Flock’s contracts and sign the certification that says they understand the “purpose” valid for each individual contract.

This is clearly unworkable; it is a recognized, and “solved” problem. Some states centralize their processing for these certifications. In those states, vendor employees can certify with the state CSA (typically state police or department of public safety), who retains their background check and information on file to share with other agencies using the same vendor.

In those states, vendor employees file a single certification with the CSA, and simply claim that they will not use it for a purpose not allowed by any of its employer’s contracts, past, present, or future, without ever seeing the contract. It’s questionable, but the FBI does not appear to have a problem with it so far.

But not all states have such a system in place. For those states, each employee needs to sign this piece of paper for each contract.

The issue is further complicated by Flock’s position that its contractual terms, which it recently altered, are negotiable and each customer can have a bespoke contract. If employees need to adhere to the terms of the contract they must, necessarily, read those contracts.

Of course, if Flock were to take the other position — that its terms are not negotiable — its contracts may qualify as contracts of adhesion, which raises its own set of problems.

Who is Not Missing

Some easy to find job titles for the folks on the list:

• UI/UX Designer & Brand Visionary
• User Experience and Service Designer
• Policy Manager (former federal prosecutor, hired from the U.S. Attorney’s Office)
• Principal Product Manager
• Manager, Solutions Engineering

There is no reason a UI/UX designer and/or brand visionary should have access to production data. This is not only a common-sense security practice, but a requirement for both SOC.2 and ISO27001 certification—both of which Flock claims to possess.

And that’s for ordinary production data; those rules apply to companies that sell caps for your ballpoint pen or that do made-to-measure T-shirts for your dog. Here, we’re talking about federally protected criminal justice information.

In any case, apparently it’s more important for a brand visionary to have access to CJI than for the Head of Security, Risk, and Compliance.^[2]

What it Means

There are only two explanations for what this list represents:

1. Flock has narrowed CJI access to 28 people — in which case several of those people have no business being on the list, Otten’s absence is inexplicable, and the company’s field technicians, Upwork contractors, and demo account users are all operating in violation of federal law; or—
2. Flock certifies everyone and handed over only a subset to make the records request go away.

Both explanations end in the same place.

Every day, Flock cameras record the movements of millions of people who never consented to surveillance and have no way to verify how their data is handled, needing to rely on Flock’s vague assurances that it is “CJIS certified.”

The CJIS Security Policy exists because criminal justice information and criminal history record information is dangerous when mismanaged. Flock’s own paperwork — the paperwork they produced to prove compliance — is the evidence that they aren’t complying.

And the certification itself? It’s a document that exposes individual signers to federal criminal prosecution for misuse of CJI.

When Flock runs sales demos on production data — real people, real movements, real criminal justice information — it’s not Flock’s name on the line. It’s the employee’s. The company that built the system, sold the system, and decided to use live data for training walks away clean. The designer who was told to sign something during onboarding risks federal charges.

Twenty-eight names. Some illegible, one conspicuously absent, and no reason to believe the list is even remotely complete. But every one of them signed on the dotted line — and not one of them is Flock.

I am not an attorney. This analysis reflects my interpretation of CJISSECPOL, contract language, and law, and is subject to change. Contracting agencies should consult qualified attorneys regarding their specific agreements.

Precrime as a Business Model

In the broader context of AI doing investigative police work—something Flock is pushing hard with Nova and its “Night Shift” feature—Langley had this to say when asked about Minority Report:

[When] you think about it, it was decades of arrests with only one wrongful arrest. How nice would that be if our current judicial system and policing system only had one wrongful arrest and multiple deaths? That sounds great.

The only problem with this system, according to Langley, is that the “terminal decision” lies with the “precog”^[1] rather than a human—the old “local decision” mantra Flock repeats in public, but puts aside when it unilaterally removes “permanent” information from its product.

Langley either completely misses the point of the movie, or he aims to bring about its dystopia in his stated, quixotic quest to eliminate crime.

Minority Report’s problem was never that the precogs were, like AI, “inhuman,” as Langley puts it. The movie is a warning that putting blind faith into a system—any system—is a terrible mistake.

The term “Minority Report” in the movie’s universe refers to an outlying data point: a piece of evidence that contradicts the other evidence and, at minimum, raises doubt about the system’s fallibility. The government’s solution in the movie? Purge minority reports from the record and hide their existence from the public.

The “one wrongful arrest”—which was actually a conviction—serves to highlight that the system has always been fallible. There are likely thousands of innocents who could not have been convicted but for the purged minority reports, removed from society “to eliminate crime in America.”

It’s a fitting reference for a company whose approach to inconvenient data is to make it disappear.

Blurring the Line Between Arrest and Conviction

In the same interview, Langley speaks on real-world problems in proving crime. The first, he claims, is that people will no longer come forward as witnesses.^[2] He continues:

The second is: our expectations of truth have gone through the roof. And and this is, like, largely a good thing, but you know, people like you and me watch NCIS on TV and we assume there’s cameras everywhere. … and you watch [shows like NCIS] and you’re like, “Oh, like this is how [it works]”, but the real world doesn’t work this way.

And so you you get to a judge, you get to a jury, and absent incredibly hard evidence, an arrest will not occur. And that’s actually, I think, good. We’re holding ourselves to a higher standard of eliminating wrongful arrest, but that kind of moves the difficulty level up. And then those two things are compounded by [the third issue of] a staffing crisis, right?

There’s a tell buried in this quote: Langley keeps saying “arrest” when he means “conviction.” He did it with Minority Report, and he does it again here. Judges and juries don’t decide arrests—they decide convictions. The standard for arrest is probable cause, which is far lower than the courtroom standard of beyond a reasonable doubt.

This conflation is not accidental. It’s strategic. Flock’s product is strong enough to generate arrests—point a camera at a road, flag a plate, send a cop. But generating a conviction requires evidence that can survive cross-examination, expert challenge, and judicial scrutiny. As we’ll see below, Flock’s evidence authentication doesn’t clear that bar. Langley blurs the terms because admitting the distinction would expose the gap between what Flock can trigger and what Flock can prove.

The CSI Effect and the “Burden of Truth”

Langley is gesturing at something real: the so-called “CSI Effect,” where jurors exposed to forensic-heavy TV dramas expect more scientific evidence than prosecutors can realistically provide. It’s a documented phenomenon, and it has made some prosecutions harder.

But Langley doesn’t frame it that way. Instead, he frames rising evidentiary expectations as a problem to be solved—a “difficulty level” that Flock can help overcome. The implication is that courts should lower the bar, or that Flock’s evidence should clear it. Neither follows. The standard in criminal proceedings exists to protect defendants from wrongful conviction. That standard hasn’t “gone through the roof.” It’s exactly where it’s supposed to be.

What has changed is that Flock wants to be the one supplying the evidence—and the evidence it produces, as we’ll see, doesn’t hold up.

Why Flock’s Evidence Doesn’t Hold Up

First, a caveat: I don’t claim authority on the Rules of Evidence. It’s a complex topic. If any lawyers want to correct me on anything, please reach out.

Here’s the gist: you can’t just make shit up and throw it at the judge. Courts require any evidence to have a basis and to be introduced by someone. This is, in part, why the prosecutor can’t show up with bodycam footage of you rolling up to the Louvre with your ladder—a police officer who was wearing the bodycam has to show up and say “I saw this guy carrying a ladder through the streets of Paris.”

When it comes to Flock footage, that means either (1) a witness comes in and says “I saw this,” (2) an expert comes in and says “this is authentic and has not been tampered with,” or (3) the court relies on more circumstantial evidence like metadata and affidavits.

Option 1 is impossible—there is no witness. Option 2 is expensive and exposes technical details in open court. Which leaves option 3: the weakest possible basis and, apparently, the focus of Langley’s complaint about standards going “through the roof.”

How Flock Authenticates Evidence

The details are sketchy, because of Flock’s continued lack of transparency, but I believe that some time last year, Flock changed how it authenticates evidence. Where it used to sign an affidavit on request, it now appears to use an automated process. Based on what I can determine, this is roughly how it works since July 1, 2025:

• A Flock camera takes a picture
• It creates a hash (shortened representation) of the image
• Flock stores the image and the hash
• An investigator goes into the Flock portal and downloads an image
• (Optional) Flock deletes the image due to retention periods, but keeps the hash
• The investigator, months later preparing for court, uploads the stored image to Flock
• The server generates a hash of the uploaded image
• The server compares it to the hash stored for the original capture
• The server returns a PDF with the image and the date, time, and location of capture that says “we checked: we took this picture and these items all belong together.”

Sounds reasonable. It isn’t.

The Chain of Custody Problem

The Chain of Custody is a key part of the rules of evidence: you have to be able to show that evidence has not been tampered with. For physical evidence, there are rigid protocols—sealing, unsealing, signing in and out of secure storage.

For Flock’s images: nobody, most likely including Flock, knows who has had access to the image, the metadata, or the hash at any point in the process. This is the reason CJIS requires permanent, immutable audit logs.

Images captured by Flock cameras are stored unencrypted on the device before transmission. Flock has previously said images are stored “for up to 7 days” on the camera, which means the metadata—capture times, location data—is also stored for up to 7 days.

This asynchronous processing is a technical necessity when operating over spotty LTE networks, but it also means there is a multi-day window in which images and metadata sit on an unattended, unsecured device.

It’s the digital equivalent of finding a dead body in an alley and saying “we’ll come back in a few days to collect the evidence.”

The Metadata Integrity Problem

There is no mechanism to validate that the metadata belongs to the image. A properly secured device would have a TPM (Trusted Platform Module) that cryptographically binds the data, image, and hash together so they cannot be separated, altered, or accessed independently. Flock’s cameras are not such devices and, from the teardowns I’ve seen, contain no TPM.

By all appearances, the file hash and the metadata are simply stored in AWS alongside everything else. Anyone with access to AWS—a Flock employee, a compromised account, a contractor—could update the data. With a few keystrokes, a photo of your car taken in Langley, VA, could be associated with a camera in Paris, IL.

Flock’s automated system will attest to this fact in court.

What Flock’s System Actually Proves

Flock’s authentication is a convoluted version of “trust me, bro.” Instead of verifying that a photo was taken where it was taken, when it was taken, it attests only to the fact that an image downloaded from Flock matches another image in Flock’s system.

That’s not authentication. That’s “both our watches say it’s 2:37pm, so it must be 2:37pm”—while ignoring that you left them unattended in your hotel room for three days before driving from Chicago to L.A. All it proves is that both watches show the same time.

Langley wants us—and the courts—to accept 2:37pm as the absolute, indisputable truth. The times match, and he’s wearing one of the watches—how could it not be the truth?

The law demands more, and so should the courts when they are deciding someone’s life and liberty.

Staff shortage be damned.

FOIA mode is now available from the theme selector in the top right:

Enabling FOIA mode will style the page in … well, FOIA mode:

In FOIA mode, an advanced simulator will simulate printing the page in black and white, it will apply redactions at random, without legal justification, before finally simulating scanning it at an angle to prevent successful character recognition and text extraction.

It is certainly the most inefficient way to render data, but it appears the one the government prefers; Roma locuta, causa finita.

You’re welcome.

FOIAing the FOIAers

haveibeenflocked.com retains no data on its regular visitors, but the privacy policy does permit collecting and retaining data on government visitors.

Although accessing the public records maintained on your local police department’s transparency portal is impossible when you’re using a VPN, anyone is welcome to visit this site using one. In fact, the Houston HIDTA bulletin that was forwarded by the FBI notes:

Law enforcement should be cautious when accessing haveibeenflocked.com, as it is unknown what information the site may be collecting. It is recommended that this website not be accessed by any computer that is connected to a law enforcement agency network

Below are the aggregate numbers of requests from government networks since about two weeks after that bulletin was forwarded to the broader “intelligence community” by the FBI.

The announcement is, of course, almost entirely marketing gibberish; to the extent that there’s any meat on its bones, this is it:

Bishop Fox’s offensive security experts will engage in complex, multistage and multilayer adversarial testing across all of Flock’s products, both hardware and software. The results and any ensuing updates will be communicated transparently to reinforce confidence in Flock’s strong security posture. — Dumb title omitted, Flock via GlobeNewswire, Feb 2, 2026

The key phrase is “multistage and multilayer adversarial testing.” This suggests^[1] they will engage Bishop Fox for some good old-fashioned red teaming.

It would be good news. If this announcement passed the smell test.

Why would Flock announce this to the world?

For readers not familiar with the term, “red teaming” is industry shorthand for hiring people who will attempt to break into your systems. It can include everything from physical entry (by breaking into buildings), to social engineering, to “hacking” systems over the Internet.

This is an incredibly useful exercise for security teams. Red teams (and actual attackers) can test vectors that employees typically can’t—for example, leaving USB sticks containing malware in company parking lots, putting on a hi-viz vest and carrying a clipboard into the server room, or sweet-talking Sam from HR into giving them an employee login.

If you want to test your everyday preparedness, announcing it to the world is not a good idea, for obvious reasons.

Perhaps more importantly for a company like Flock, an announcement like this sets expectations:

The results and any ensuing updates will be communicated transparently to reinforce confidence in Flock’s strong security posture.

A red team exercise at any organization, let alone one with a track record like Flock’s, is not a one-shot third-party validation exercise. Testing and addressing vulnerabilities is a months-long coordinated effort between senior management, in-house security staff, external consultants, and engineering teams.

If done right, the result is not a report to be presented in a shareholder call; it’s a binder documenting the work your management and engineering teams will be doing for the next six months.

And that’s just about the best-case scenario, which I do not expect for Flock.

Instead, Flock blasts out a press release with little to no context, creating unnecessary friction between shareholders who simply want a stamp of approval and security teams who want meaningful, long-term fixes.

Flock uses the language of success to set itself up to fail.

The Bishop Fox Choice

Bishop Fox is generally a well-regarded offensive security firm—the kind of company you hire when you’re serious about finding vulnerabilities. But …

Late last year, I published “Y Combinator funds both surveillance infrastructure and the machinery to silence its critics” which described some of the Y Combinator ties between Flock and its alleged other cybersecurity partner, Cyble.

I say “alleged,” because in that post, I questioned how formal the relationship is, writing “I would expect one of them to do a press release announcing a ‘strategic partnership.’” Here, Flock did not choose Cyble. It chose Bishop Fox. And it put out a press release.

While the ties between the companies do not suggest the same level of intertwinement as between Flock and Cyble, interesting overlaps remain.

Reddit co-founder Alexis Ohanian was Flock’s first investor while Reddit co-founder Steve Huffman currently serves on the board of Bishop Fox.

Chris Castaldo, Flock’s new CISO, worked at IronNet CyberSecurity before being hired at Flock. Don Dixon, managing director of Forgepoint Capital, serves on the board of both IronNet and Bishop Fox.

Castaldo also worked with Will Lin—another managing director and founding member of ForgePoint Capital^[2]: they co-founded the non-profit Security Tinkerers in 2018 and continue to collaborate on it today.

The “follow the money” connection between Flock and Bishop Fox is not as obvious or direct as the one between Flock and Cyble, but the close personal relationship Flock’s new CISO maintains with managing directors and board members of a “neutral third party” that could add or remove billions from Flock’s valuation raises serious red flags.

We’ll see if Flock publicly acknowledges this appearance of a conflict at any point before the “results” are in, or if we’re expected to take everything at face value.

The impossible bind

The CJIS Security Policy (CJISSECPOL), also name-dropped^[3] in the press release, creates an inescapable problem for any “production” testing of Flock’s systems.

There is a commonly-used CJISSECPOL workaround for giving contractors temporary access to CJI without full vetting: a Flock employee “escorts” the contractors while they work. This avoids fingerprinting, background checks, and the cascade of compliance certifications that would otherwise be required from every agency customer in states without centralized contractor vetting.

The problem is that in an “escort” scenario, the escort is legally required to prevent Bishop Fox from accessing unencrypted CJI. CJISSECPOL § 5.1.1.5 is explicit: physical access must be “controlled” and the escort must maintain “observation” to ensure the contractor cannot view protected data.^[4]

As soon as Bishop Fox successfully discovers a vulnerability that exposes real data—which is, after all, the entire point of red teaming—the escort has failed in their duties. The incident becomes reportable under CJISSECPOL. CGAs must be notified, as well as the FBI, and mitigation plans must be submitted.

Success equals failure. The very act of doing the security audit competently would trigger mandatory incident reporting.

Neither alternative works

For Flock, as the defending “blue team,” there are two paths forward, and both lead nowhere good.

Option 1: Test on a replica environment.

There is a lot of evidence of Flock using development-specific code and keys in production and vice-versa, suggesting poor logical separation and cross-environment contamination. If I had to make a list of “organizations I would expect can roll out an accurate replica of their production environment,” Flock would definitely not be on it.

Even assuming Flock could create an accurate replica software environment, if your penetration testing is multilayer and includes physical security, you have to include the security of your office and server buildings, as well as any parts of your network you’re leaving unattended on a stick on the side of the road.

And while a replica might yield valid results for a blue team interested in making improvements, because we can’t verify the fidelity of the replica, it would invalidate a lot of the “third party” claims that Flock raises in its press release. (Again, I ask: why announce it in a press release?)

Option 2: Test in production.

This creates the impossible bind described above. But even setting aside the escort paradox, testing in production without the escort workaround would be worse.

Some states, through their CSAs, have centralized vetting for contractors. Many do not. For states without centralized vetting, each Bishop Fox employee with access to unencrypted CJI would have to be fingerprinted, background checked, and certify their knowledge of, and agreement with, CJISSECPOL to each Flock customer with an active CJIS security addendum.

To be compliant with CJISSECPOL, all governmental Flock customers in those states must independently ensure this has happened. Failing to do so, even in a single jurisdiction, would bring all of Flock’s customers—including those in states with centralized compliance—out of compliance the moment Bishop Fox touches a live packet.

We already know Flock sends data to Denmark and the Philippines. The certifications I have received in open records requests did not include these contractors.

Ultimately, it’s on local criminal justice agencies and their state CSAs—not Flock—to remain in compliance with CJISSECPOL.

If Flock were to add another subcontractor to access its customers’ CJI without obtaining necessary authorizations, conducting the necessary background checks, and providing the required compliance documentation, it would bring its agency customers even further out of compliance.

We’ve tried nothing, and we’re all out of ideas

Flock continues to sit on the report by GainSec, which documents dozens of vulnerabilities that were reported to Flock in February 2025 but, by all accounts, remain unfixed. It also continues to ignore the unrelated issue from late 2025, where it hardcoded passwords in production.

The red team should have no trouble finding and flagging these issues. Then we’ll have another report for Flock to fail to act on.

In its press release, Flock writes that “[t]he results and any ensuing updates will be communicated transparently to reinforce confidence in Flock’s strong security posture.”

Flock could start on that today by acknowledging and fixing the already-documented vulnerabilities in its products.

Flock could also own up to all the security incidents it has experienced, from accidentally disclosing a file with customer emails, to hardcoding passwords in roadside cameras. It could transparently implement fixes, or even provide a schedule for these fixes.

The company could address the issues with compliance, which include failures to mitigate critical security vulnerabilities within 15 days as CJISSECPOL requires, designing the system to disseminate CJI indiscriminately, and leaking entire murder investigations.

Instead of falsely claiming “we have never been hacked” and removing accountability measures, Flock could work with independent security researchers, rather than try to get them to sign NDAs.

Flock could even work with CSAs and the FBI, which are authorized to audit Flock’s systems. After several unanswered requests to the Iowa Department of Public Safety (Iowa’s CSA), in December 2025, I even requested the FBI perform such an audit, citing incidents where Flock disseminated warrant information from NCIC, and the 157 pages of murder investigation mentioned earlier.^[5]

The company has not issued a single press release indicating it has done, or plans to do, any of these things.

From inception, this announcement has all the hallmarks of compliance theater—perhaps producing a meaningless report by an “independent” third party, before CJISSECPOL’s stricter “Supply Chain Risk Management” controls come into full effect with version 6, is a way to avoid the Department of Justice needing to wade into the mess Flock, local agencies, and CSAs have created.

Flock’s goal should be to improve its security posture, not to “reinforce confidence” in it. One is security, the other is managing public perception—i.e., marketing.

We are aware that agencies across the country, particularly in states with broad public-records laws, are seeing increased PRA/FOIA activity seeking, among other things, LPR search logs.^[1]

Recently, a third-party website began aggregating search information that appears to have been released through these public-records processes. We recognize that seeing investigative search activity displayed publicly can raise understandable concerns about officer safety, investigative integrity, community perception, and compliance with state law.

@“What you Need to Know About Recent Online Disclosures”, December 8, 2025

Yes, Chris. Those concerns have been the point all along.

He promises new tools for 2026, and continues:

Until these new [redaction] tools are available, we recommend: Reviewing your sharing settings and considering a temporary shift from Nationwide Lookup to Statewide Lookup, if that better aligns with your agency’s legal guidance and operational risk posture.

This recommendation—and this plan to bury the bodies—is an outright admission that the Flock system is so critically flawed, and so widely abused, that it cannot survive public scrutiny.

Of course, activists everywhere commend the recommendation to leave the nationwide network. Flock has spent years claiming this network is essential for public safety. Will we have anarchy in the streets, or will it turn out to have been optional all along?

Anyway. Let’s talk about hiding the evidence.

Officer safety and investigative integrity

First, let’s define what we’re talking about with these terms.

Many states have statutes similar to this one from Texas:

“Criminal history record information” means information about individuals collected by criminal justice agencies consisting of identifiable descriptions and notations of arrests; detentions; the filing of complaints, indictments, or informations and dispositions arising from complaints, indictments, or informations; sentences; correctional status; and release. It includes identification information, such as fingerprint records or photographs

A person commits an offense if the person knowingly or intentionally obtains criminal history record information … uses the information for an unauthorized purpose, or discloses the information to a person who is not entitled to the information.

This would appear to cover most of the information in this search log entry:

The only item that was redacted by Flock and Houston PD was the license plate. This person’s name, date of birth, phone number, and criminal history (“AGG ASSAULT DEADLY WEAPON”) were all entered into an insecure system and then fired off to 3,000+ agencies.^[2]

Not Houston. Houston hides who it shares data with.

In the statute’s terms: the information was knowingly or intentionally disclosed to 3,000+ persons not entitled to the information. Not by the officer entering the information, but by Flock—an unauthorized person that first knowingly or intentionally obtained the information, and then re-disclosed it.

This continues to happen, even after a 157-page multi-jurisdictional report on an active gang-related murder investigation got leaked, and, just yesterday, a LINX report on a fatal hit-and-run surfaced after it was pasted into a case number field.

CHRI, CJI, intelligence data … various terms and statutes apply depending on the exact information, context, and state. But the general pattern holds: statutes criminalize indiscriminately sharing this type of information.

So … might I be so bold as to suggest not indiscriminately sharing such information? Not with Flock, and not with the nationwide network.

I’m not a cop, but I think not criming could also help with the whole “reducing crime” thing.

Just a thought.

Compliance with state law

While I somewhat enjoy the idea that snarky blog posts and reposting audit logs proximately cause police to violate the law, I’m sure that’s not what Chris means.

The only possible “concern” he could be referring to is the concern that this website would be publishing evidence that agencies are not compliant with state law. This is a valid concern. See above.

The solution is the same as before: stop breaking the law.

A pattern emerges.

Community Perception

And finally, the only concern worth taking seriously: “community perception.”

This is valid.

The community will not perceive it as a positive that the system they were promised would be used to find missing children, is actually being used for reasons like “Operation Homeless Intel- Make your own case!”, to perform blatantly unconstitutional searches, or to investigate “gypsy crap”.

The community may also take a dim view of Flock users obtaining a 30-day history of someone’s whereabouts simply because they were “‘suspicious’ per a karen.”

Flock and police promised the community these things would not happen.

They promised “the system requires a detailed search reason” and that these “detailed reasons” would be stored in a “permanent log file.” Police chiefs stood before city councils, assuring them logs would be audited regularly. Many even adopted policies to that effect.

Yet, this website shows nationwide searches for “inv” and “sus” have consistently made it past the scrutiny of every single police chief who promised the community that “we own the data and we have full control over who gets access.”

There have been hundreds, if not thousands, of these promises. ALPR policies that require audits are everywhere. Every day, these policies collectively promise, police officers across the country will audit hundreds of thousands of searches.

Every single “sleepy Virginia town” that commits to reviewing nationwide search logs is committing to reviewing around 150,000 searches every single week. In each of these audits, every single instance of “sus” or “inv” should—ironically—be considered suspicious and be investigated.

Even assuming only a fraction of Flock’s customers commit to audits, on average there should be approximately two million phone calls or emails between police departments demanding answers about “inv” and “sus” every single day.

I have not heard of one.

Hundreds of thousands of entries containing only "inv", "sus", or "test". Or, you can “Make your own case!” in "Operation Homeless Intel".

“Agencies own and control the data” (but “for clarity,” that’s not in the TOS).

And now, the “permanent, immutable audit logs” are on the 2026 roadmap to redact.

Community perception is a valid concern when such a massive breach of trust is put on full display.

Redaction: Why not?

I want to touch on redaction. It’s not complicated. It’s not expensive. This website’s entire redaction library, which is relatively complex because it aims to balance privacy with transparency, comes in at less than 1,000 lines of code. It takes seconds to process millions of records on a small laptop.

It takes even less code to truncate a field so you can’t paste entire reports in there.

For any non-programmers who might be reading this, this is literally, without exaggeration, the single change needed, and the one Flock has failed to implement for years now:

<input type="text" />                  <!-- ❌ active murder investigations are compromised -->
<input type="text" maxlength="420" />  <!-- ✅ professional-grade solution engineering -->

That’s it. maxlength="420".

420 characters is more than enough to accommodate the typical “reason” length of 7 characters. Even if a cop decided to go a little cray-cray and write an actual justification, 400 characters would accommodate it. They can use the remaining 20 for their badge number.

You’re welcome, seven-billion-dollar tech company.

You can pull my address from Flock Nova and mail me a check.

But instead of making this simple change, Flock hangs on to its unrestricted fields. There is not even a “hey maybe don’t enter this,” like there is for searches that might more overtly violate law.

Now, when these problems are finally coming to light, Flock’s solution is to go into the archives and ham-fistedly replace entire categories of data—which its customers need to be able to access to comply with state, local, and federal law (not to mention the ALPR policy promising audits)—in its “immutable” and “permanent” logs and replace them with the word “redacted.”

Right now, this change appears to affect existing records. Altering public records is a criminal act in many, if not most, states. If Flock is actually making the information available to police and police are ticking (or leaving ticked) the “hide information” boxes, they may run afoul of open records laws. Either way, it looks like someone is about to take on some extra liability (hint: Flock added a new paragraph to its ToS on December 19—guess what it’s about?).^[3]

Flock is handing police a shovel to bury the records while quietly changing the contract to say the legal fallout is not on Flock.

The solution Flock is implementing is overly complex and costly. Scroll up. Does that need to be delivered in “Early 2026,” after the lawyers have had time to review and deploy the TOS change, or could it have been implemented before Colwell’s email was even done sending?

Who needs redaction when you have AI?

The reason why Flock will not redact anything is its business model. It is causing the company’s constructs to collapse.

Flock does not, as it likes to pretend, exist only in the abstract. Its cameras aren’t only found along country roads, magically snapping pictures of “only license plates” in its stated effort to eliminate crime.

Flock doesn’t exist to eliminate crime. It exists to make money. It has a fiduciary duty to its shareholders to prioritize profit. To make a profit, it requires unregulated data, free and clear, unencumbered by public records law or restrictions on criminal justice information.

It requires data it can feed into LLMs. Data it can mine. Data it can convert into “actionable insights” and then sell. It doesn’t matter how the data gets there, what the quality of that data is, how secure it is or what it’s being used for.

Amassing vast amounts of data is a substitute for expensive manual training that often requires sending your data abroad. And what better way to gather that data than convincing the government to pay you to pepper public places with cameras and microphones? What better way to be exempt from permits and licensure?

These days, everyone has AI. It’s the data that differentiates. It’s the data Flock has seized for its own use. Colwell knowingly misleads his customers when he says “all Flock data is owned and controlled by the agency that collected it.”

Flock’s customers don’t collect data when they use Falcon/LPR. Flock does. This isn’t pedantry: it’s language in a mass email from a high-level Flock executive about legal compliance.

Flock’s terms of service state, “for clarity”:

Customer Data does not include the underlying raw Footage captured by the Flock Hardware […] “Footage” means still images, video, audio, and other raw data captured by the Flock Hardware or Customer Hardware via the Flock Services.

So instead of addressing any of the actual problems, which would mean collecting less data or deleting some of the data, Flock’s solution is to hide the data from its customers.

The Sunlight Phase

When I launched this project, I included a statement on the front page titled “Should you be publishing this information?” It argued that sunlight is the best disinfectant.

Flock’s recommendation to shut down the nationwide network proves it.

This website aggregates and reformats already-public information. This information represents a fraction of what’s being shared with Flock and its government, commercial, and private partners on a daily basis.

Policies exist to prevent the release of this information—they are not adhered to. Laws and regulations exist to enforce the policies—they go unenforced. Police, Flock, and politicians have been ignoring these problems for years while your private movements continue to be collected, catalogued, sold and traded.

This website exposes the problem because, as the old saying goes, sunlight is the best disinfectant. Law enforcement and legislation are needed to address the cause of the problem, and we highly encourage you to bring this site to the attention of your legislators.

We believe mass surveillance has no place in a free society, and this data should not be collected to begin with. If it is collected, warrants should be used, lookups should be rare, and police and private parties, like Flock and HaveIBeenFlocked.com, should not be permitted to act without functional restraints or oversight.

We have entered the sunlight phase.

In his email, Colwell writes that the new “enhancements” are planned for early 2026. That gives them about 48 hours to find a new way to hide the truth.

We’ll be here. Filing open records requests.

Is it misleading to claim Flock LPR devices use encrypted storage when they don’t?

Is it misleading to tout “CJIS compliance” when you know that’s entirely different from CJIS validation—actual verification, by the FBI, that your system meets federal security standards (like using encrypted storage)?

From: Garrett Langley <noreply@flocksafety.com>
Sent: Monday, December 8, 2025 3:37 PM
Subject: Fact Check: No Hack. We will never stop fighting for you.

Hi VA - Staunton PD Team,

I’m writing to you directly because I want there to be zero confusion about what’s happening.

Flock has never been hacked. Ever.

Flock is CJIS compliant.

Flock does not share, or resell your data. Nor have we ever.

Flock adheres to the highest security standards, including: NDAA, SOC2 (Type II), SOC3, ISO 27001, HECVAT, FERPA [and “aligned with” NIST and CAIQ].^[1]

Flock is building tools to help you fight the real crime affecting communities across the country.

Many activists don’t like that.

Let’s call this what it is: Flock, and the law enforcement agencies we partner with, are under coordinated attack.

The attacks aren’t new. You’ve been dealing with this for forever, and we’ve been dealing with this since our founding, from the same activist groups who want to defund the police, weaken public safety, and normalize lawlessness. Now, they’re producing YouTube videos with misleading headlines. They’re also trying to turn a public records process into a weapon against you and against us.

Make no mistake, we’re fighting this fight for you, and, I hope, with, you. I remain committed to building world-class technology to help you keep your communities safe. And doing so in a transparent, secure, and privacy centric way.

Garrett

Police Chief Williams, a 41-year veteran, responds:

As far as your assertion that we are currently under attack, I do not believe that this is so. I have dedicated the last 41 years of my life to serving the citizens of the City of Staunton as a police officer, the last 22 as the police chief. What we are seeing here is a group of local citizens who are raising concerns that we could be potentially surveilling private citizens, residents and visitors and using the data for nefarious purposes … In short, it is democracy in action.

Chief Williams concludes by stating he trusts Flock will continue to supply “state-of-the-art” systems to make Staunton safer.

Langley responds:

Thank you sir. … And understood on the attack perspective. It’s tough every day waking up to stories online that are misleading and only represent one side of the story. It’s[sic] sounds like in Staunton you’ve had a civil and orderly discussion. That is, in fact, democracy in action! Sadly, other cities have not been as well served.

Criminal justice information has leaked through Flock’s system before, and continues to leak. Flock claims it hasn’t been hacked, but they themselves wrote the code to record customers and transmit the recordings overseas. Flock simply hands over the the data to unvetted foreign contractors.

It should be obvious by now that the problem isn’t that the story is onesided: the facts are.

To prevent more unfavorable facts coming out, Flock continues to actively fight transparency and tries to shut down websites already publishing facts.

“They’re producing YouTube videos with misleading headlines.”

No, Garrett, your system is broken. If it were CJIS validated, it would be straightforward to publish the documentation. Each of your customers would already have a copy.

They are boilerplate public records, not state secrets; why not release them to “set the record straight”? Why not provide hard evidence that the system is secure? Not to appease activists but to strengthen public safety—prove to your customers that their data is secure and that it complies with federal law.

Instead, Langley says citizens calling for security, transparency, oversight, and accountability is an “attack.” And that such things “normalize lawlessness.”

Williams calls it what it is: democracy.

Publish the validation documents. Let the facts favor you for once. That’s democracy in action.

• Complete email exchange (PDF)

Their argument last month was remarkable: allowing non-government employees to access ALPR data would “violate the spirit of the Fourth Amendment.”

To support this claim, the cities borrowed key talking points from the anti-surveillance crowd, including that collecting and accessing detailed location information about anyone creates tremendous potential for abuse. They cited examples like suspicious spouses tracking partners and raised hypotheticals about disgruntled citizens tracking government employees.

What the cities don’t mention is that Flock has no special rights under Washington law. They’re not a law enforcement agency, or even a government agency. The system doesn’t comply with laws concerning intelligence gathering or regulations governing sensitive information. If it did, the data would likely be exempt from the PRA.

Unlike other cloud services, which are generally protected by federal law, nothing prevents a divorce attorney from subpoenaing records about a potentially unfaithful spouse from Flock. Sedro-Woolley and Stanwood propose that we remove the only remaining, albeit woefully inadequate, safeguard that at least sometimes catches disgruntled Flock employees or customers stalking citizens, or accessing data despite assurances to the contrary.

The cities are right about one thing: the potential for abuse is enormous; giving Flock the keys to the kingdom is a terrible idea; and, yes, the whole system does violate the spirit of the Fourth Amendment.

The cities’ conclusion, however, is dead wrong: you can’t prevent abuse by hiding it. It must be addressed at the source.

• The full brief (10/8/2025)
• The defendant’s response (10/8/2025)