Federal Court Says Three Hits From Flock ALPR Cameras Isn’t Enough For A ‘Carpenter’ Violation

The Supreme Court’s Carpenter decision created a warrant requirement for obtaining location data from service providers. It was a limited ruling, albeit one that has had far-ranging implications.

Thanks to this ruling, law enforcement agencies have started buying location data from third-party brokers, rather than suffer the apparent indignity of having to ask a judge to approve a warrant. The underlying theme of the ruling — that the Fourth Amendment ain’t what it used to be now that everyone’s online all the time — has seen it applied to cases where location data isn’t the underlying concern. Anything law enforcement might use to engage in tracking of individuals is now under additional scrutiny.

And while this is mostly a welcome development, it also means there’s still a lot of unsettled law that isn’t going to get settled quickly, because the confines of the ruling and the limited upheaval of the Third Party Doctrine are often in conflict.

But some cases are easier than others, even if they’re actually still a bit complicated. A recent ruling handed down by a Virginia federal court applies Carpenter to automated license plate readers. It’s not the first to do so but it might be the first that involves Flock Safety, a manufacturer of license plate readers that’s equally willing to sell to both private and government entities.

Flock first made a name by hawking its wares to gated communities and HOAs, preying on their desire to keep outsiders out and insiders subject to the restraints of their HOA agreements. Then it realized an unlimited revenue stream existed in the public sector and started courting cops. And then it realized it could leverage both of its markets by encouraging the private sector to make its ALPR shots available to cops without the friction of subpoenas or warrants.

And that’s why this decision [PDF], coming to us via FourthAmendment.com, is especially interesting. It’s not all that interesting that a court might decide the use of three ALPR/vehicle photos obtained from Flock ALPR cameras might not meet the “long-term location tracking” standard loosely applied by the Carpenter decision. What’s far more interesting about this decision is its discussion of how Flock and its network of cameras operates, especially when assisting law enforcement during investigations. It also highlights aspects of Flock camera operation users and non-users may not have been aware of.

Here’s how this worked out for robbery suspect Kumiko Martin Jr., who sought to have the Flock ALPR images obtained by investigators suppressed:

The Court is cautious to not hinder law enforcement’s use of modernizing surveillance capabilities in the public sphere lest the Court “embarrass the future.” Carpenter, 585 U.S. at 316 (quoting Nw. Airlines, Inc. v. Minnesota, 322 U.S. 292, 300 (1944) (internal quotation marks omitted)). This Court must rule on the facts as they are and may not speculate about what the future may hold for Flock’s capabilities. Today’s ruling is limited to the facts of this case as they are at the time of this ruling, including the limited number of Flock cameras in the Richmond area and the limited number of pictures taken of the exterior of Martin’s vehicle. Accessing Flock’s database, which captured only three photographs of Martin’s vehicle during the relevant 30-day period, did not allow law enforcement to track or monitor the “whole of [Martin’s] physical movements,” and therefore was not a search under the Fourth Amendment.

As the court states, this finding is limited to these facts. Future cases involving the same tech may alter the Fourth Amendment equilibrium, but for now, applying Carpenter doesn’t turn three images into long-term warrantless tracking.

But there’s a lot more information in the decision, which makes the entire thing worth reading. The first thing worth noting is that Flock’s ALPR offerings don’t limit themselves to identifying license plates. The tech does what it can to flag anything else distinctive about the vehicle and add that to its searchable database.

Unlike ALPRs, photographs by Flock cameras are uploaded in full to a Cloud database that records and stores the captured data. This searchable data includes the photograph’s date, time, and location as well as the vehicle’s license plate {and absence temporariness, or obstruction thereof), the plate’s state- and/or country-of-origin, body type. make, model, color, and other “unique identifiers” such as visible toolboxes, bumper and window stickers, roof racks, and damage to the exterior of the vehicle. Flock updates the software to provide additional metrics for use in querying and reviewing the database.

In this case, the Flock camera was mounted on a pole. This camera cannot be remotely controlled, so its operation was completely passive. Even so, its limited functions still provided more info than most ALPRs currently deployed solely by law enforcement agencies.

And while most ALPR cameras will capture at least some of a vehicle’s exterior, searchable databases of plate records tend to be limited to plate/location/time/date info. Flock’s software, however, basically allows investigators to perform reverse image searches of almost any distinguishing feature in hopes of finding hits based on vehicle description, rather than license plate numbers.

[W]hen Officer Redford queried the Flock database for vehicles that matched the description of the Dunston Robbery suspect’s car obtained from the Valero cameras, the Flock system limited its results to the 30 days preceding April 22, 2023.

This time limit is not specific to Flock. Retention periods can be indefinite. But ALPR data retention is limited by law in Virginia to 30 days, so Flock’s software applied this limit to provide the investigator with search results within the statutory limit. Nothing stated in this decision (or by Flock itself) suggests Flock doesn’t retain these records for a longer period, no matter what limitations might be placed on its government customers.

Officer Redford’s query returned 2,500 results (photographs), which is the maximum that Flock’s system shows per search. Officer Redford then manually reviewed those 2,500 pictures and found two of the suspect’s vehicle, which Officer Redford was able to identify based on the unique stickers Unlike the Valero security-camera footage, the Flock pictures also identified the vehicle as gold with a Virginia license-plate number of UAL-6525.

Given this option, it’s unsurprising law enforcement agencies are the fastest-growing market for Flock ALPR readers. This makes regular ALPRs look positively anemic in terms of technical capability. Unfortunately, it also encourages officers to do things that might be constitutionally unwise, like working backwards from a vague vehicle description to a supposed suspect by performing underinformed and ill-advised searches. It’s the same thing we’ve seen gone wrong with facial recognition tech. Now, we’re not only involving possibly innocent vehicles, but the drivers of those vehicles who will be treated as dangerous criminal suspects just because the software applied the wrong tag or an investigator leapt to conclusions not actually supported by the search results.

To be fair to Flock, it’s at least trying to curb some of the collateral damage.

The cameras are not designed to capture pictures of humans but may do so incidentally. If that occurs, however, the database does not allow searches based on biometric or other human-based characteristics that would allow law enforcement to scan for individuals.

But that’s not exactly good news. The capability to utilize these ALPRs for facial recognition is already at least partly there. At some point, Flock is going to decide it’s worth the PR hit to turn its cameras into ALPR/facial recognition hybrids, instantly doubling the issues created by its system’s ability to flag cars by description, which is far less distinct than limiting search results to license plate numbers.

Also concerning: the “share” option built into privately-owned Flock cameras doesn’t limit itself to local law enforcement agencies. Agreeing to share image captures with law enforcement means officers anywhere in the county can access data or recordings made by these cameras. This decision notes that the officer had access to 188 public and private cameras in the surrounding three counties alone.

Given what’s been put on the record about Flock cameras and their capabilities, there may be a solid case in the future that might invoke Carpenter’s warrant requirement. Unfortunately, there’s a much larger hurdle to be cleared here. Cameras only capture images of cars on public roads (at least in most cases). They do not track people’s movements outside of their vehicles. So, it’s unlikely any court will find this sort of post facto “tracking” a violation of the Fourth Amendment. But the added capabilities — and the government’s reliance on privately-owned cameras when performing searches for vehicle movements — might pose an issue later on when more precedent involving unforeseen technological developments is generated. For now, at least we know a bit more about how Flock operates and how investigators are using it to search for cars, rather than plate numbers. It’s informative, to be sure, but it’s not going to help this suspect avoid a trial on robbery charges.