Millions have witnessed the incident involving Flock cameras tracking me while I was in a $155,000 Range Rover, followed by four police officers approaching my wife and me in a parking lot. Human mistakes played a role, but Flock’s AI technology amplified these issues at an alarming speed.
In the latest edition of The Drivecast, Flock Safety Founder and CEO Garrett Langley joined me for an honest conversation about my experience and the experiences of numerous others over recent years. We discussed Flock’s vision for the future amidst the public’s scrutiny regarding privacy and Fourth Amendment issues, the increasing calls for regulation, and his previous statement labeling those who damage or vandalize Flock equipment as “domestic terrorists.”
Daily, there are reports of law enforcement misusing Flock’s technology to monitor individuals, which can lead to arrests and incarceration, misinterpretation of events, disputes in city council meetings, and assaults on automatic license plate readers—cutting them down, shooting them, or covering them with trash bags, among other actions. This has become a hot-button issue, and somehow, I found myself right in the thick of it.
New here? The Drivecast is The Drive‘s weekly podcast that gives you an insider’s look at the major controversies, stories, and personalities shaping the automotive world today. With access to The Drive‘s exclusive reporting, insights, and updates, The Drivecast strives to make everyone feel like part of the conversation.
Tune into The Drivecast on Spotify, Apple Podcasts, or Amazon Music. Enjoying it? Want to support? Leave a five-star review on your platform to help increase The Drivecast’s visibility. Have feedback, ideas, or a question? Reach us at [email protected]—we review every email.
Complete Transcript
Joel: Garrett, I appreciate you coming on the podcast. I know your schedule is packed and there’s major controversy surrounding Flock right now. You’re balancing a full life, including family, while traveling frequently. Let’s dive in; usually, I have a relationship with our podcast guests, but this is the first time we’re meeting. Given the podcast, I need to know: Are you into cars? What’s your take on driving? I’m genuinely curious.
Garrett: I hate to disappoint your listeners, but I’m not really a car enthusiast. I appreciate F1 from a technical perspective; it’s incredible how they optimize every detail to maximize performance. Funny enough, until recently, I was still driving the same Toyota Prius from college. People often ask why I haven’t switched cars, but it serves its purpose perfectly. I even modified it to have Apple CarPlay, but truthfully, I get a lot of ribbing from car-loving friends for not being more into it. With three kids, a wife, and my business, I’m trying to prioritize simpler things in life.
Joel: I get that. I wasn’t sure what you would say, considering you’re not directly involved with cars or parts supply. But a Prius, Corolla, or Camry aren’t bad options. They might not be enthusiast cars, yet they excel in fuel efficiency and reliability. The newest Prius looks fantastic, by the way.
Of course, I have an endless list of questions, and we won’t get through everything in our limited time today. We need to start with my experience, as that’s how we got onto this topic. We can’t ignore it. I’m sure you’re familiar with the details; I’ve covered it extensively myself and discussed it with your Chief Communications Officer, Josh, who’s a genuinely nice guy.
To summarize for the millions who didn’t catch the story: a human error caused a police officer to input an incomplete license plate number into the National Crime Information Center database, which is managed by the FBI and accessed by law enforcement nationwide. This incorrect information got fed into automatic license plate readers, including those from Flock, which, along with others like Leonardo and Axon, pull from that same FBI database.
The system encountered erroneous information. It was instructed to look for a license plate number, “34DTM.” It found it, but the trouble was that my press car that week—a $155,000 Range Rover—also included ‘10’ in the actual plate. Flock processed the data, focusing solely on the requested input. The local police failed to check the back-end data properly and did not confirm the actual plate number. They had a photo proving that there was ‘10’ included, but they either disregarded it or did not include it in their checks. They wrongly assumed I was involved in Grand Theft Auto, which I wasn’t.
In essence, we had human error that interacted with Flock’s technology, accelerating the spread of that mistake, but Flock wasn’t the root of the error. There was also a breakdown on the follow-through, lack of cross-referencing added layers of complications.
You can’t eliminate human error; that’s not your responsibility. I’d like to know what takeaways Flock plans to have from this situation, as Josh mentioned you spoke with the FBI overseeing NCIC and that you reached out to the LA police about the incorrect entry. What lessons will be learned from this incident? We often derive insights from such edge cases.
Garrett: First, I appreciate you laying out the facts, which indicate that many things failed. It’s easy to pinpoint the blame on “Flock did this,” but a multitude of mistakes led to this edge case. Unfortunately, you weren’t the first victim of such incidents, and as you pointed out, Joel, the key issue is the outdated NCIC database managed by the FBI, which lacks feedback loop mechanisms; it’s essentially a static comma-separated file.
We’ve attempted to work around it with a concept called “suppression.” For instance, a local agency—like Atlanta, where I reside—might see a stolen plate from California that has already been resolved as either a rental or dealer vehicle. They can choose to suppress alerts in Flock meaning, “Don’t notify us on that for a year.” This helps bypass the inability of one agency to remove it from NCIC directly.
Ideally, if several agencies—say one, two, or three—flag a tag as outdated or invalid, it should be automatically removed or at least more aggressively pushed for removal by the FBI. We maintain a cooperative relationship with the FBI, who play a crucial role in public safety, and I hope they recognize an opportunity to modernize NCIC, which is vital not just for companies like Flock but for coordinating police efforts.
Joel: I find it interesting that you noted this—especially in terms of your dialogue with the chief of police and Josh. When my incident occurred, the local police informed me that they needed to know where I planned to drive the car and warned that other police forces might act differently. They advised me that in other areas, officers might respond more aggressively, potentially showing up with their weapons drawn, unlike my experience. They mentioned that I would be safe in Plymouth since I had a case number and contact name, but travelling elsewhere could trigger alerts on Flock’s system because it relies on NCIC. If Plymouth implemented the suppression feature you referenced, it wouldn’t apply in Minneapolis.
Garrett: Correct, it only impacts Plymouth. We take the stance that decisions should remain local. Plymouth made the decision to suppress that alert, and they’re entitled to it, but that doesn’t apply in Minneapolis.
Joel: Do you think that Plymouth police quickly understood—”quickly” being a relative term—my wife would attest to this—that the only party capable of correcting the partial license plate in the system would be the officers from the unit involved? In this case, the LA Sheriff’s Department, Malibu Lost Hills station, are the only entities that can amend that issue. Do you believe that as you discuss modernizing with NCIC and the FBI, Plymouth police should be empowered to rectify the mistake directly if they consult with LA? It seems inefficient to have to navigate through so many layers to fix something.
Garrett: A potential solution would be structured like this—say you, Joel, flag a license plate “FEDER” as stolen. If I, Garrett, as the LA County officer say, “Wait, I just encountered that vehicle and it’s no longer stolen,” I should be able to indicate that it ought to be removed.
Then, I should have a certain timeframe—let’s call it a day or a week—to contest that. I think that model is much more sensible, providing the ability to propose removals while allowing us to affirm validity if necessary.
I concur that more collaboration is crucial. Similarly, I feel strongly about…
Joel: Just to clarify, when we discuss you as “Flock,” or representing the FBI NCIC in our example—who are you referring to?
Garrett: I’m positioning myself as the police department. To simplify, let’s say I’m LA County, and I register a car tagged “FEDER” as stolen. In your role as Plymouth, you should be able to question this and suggests its deletion. I, as LA County, should then have a defined period—say one day or one week—to either affirm the status or expect the plate to be deleted automatically.
Often, what happens is with rentals: a vehicle is reported stolen by the rental company and entered into NCIC. Later, the vehicle returns, indicating it wasn’t stolen after all. In our system, we have an expiration feature; entries cannot remain indefinitely. There will be a predetermined presence of 7 days, 30 days, or 90 days—timed before removal. I believe implementing a similar expiration policy for NCIC would significantly improve the management of data to prevent stale records, as you faced in your experience.
Joel: One last question in this realm. Flock employs expiration timelines, while NCIC doesn’t. In that imagined timeline I mentioned, let’s consider a situation that was avoided—say my wife opted for a manicure that day. If 34DTM had been flagged in the system and not intercepted, given that NCIC data wasn’t kept in the Flock database, how does that expiration policy function if it’s indefinite with NCIC?
Garrett: Consider Plymouth again. If Plymouth suspects that Feder’s license plate is involved in a residential burglary, they could add it to their local hot list. This would be Plymouth’s unique entry. There’s NCIC data, plus enforced expiration at the local level. It would be beneficial if NCIC adopted a similar expiration framework—let’s position it at a week or a month—so that investigators would need to verify ongoing searches regularly. This would add accountability for law enforcement and is supported by your example as it highlights the necessity of diligence in data management.
Joel: Understood. Let’s shift focus from my experience to the present and explore our direction moving forward while still touching on this topic.
Josh emphasized—and it’s prominently featured on your website—that Flock boasts a 99% success rate. I don’t want to question that outright, so in this scenario, let’s accept that as a basis. You’re averaging around—these are approximations since they’re significant figures—20 billion reads each month. With a 99% success rate, that implies a 1% error rate. One percent of 20 billion is substantial, which puts your error instances at 200 million! I was caught in one of them myself. Although it wasn’t a misread in the traditional sense, it resulted from a problematic scenario that led to an undesirable outcome.
What measures are being taken to address and minimize that 1% error rate? I mean, being at 200 million errors sounds vast—not as overwhelming given the 20 billion reads, but still significant.
Garrett: Your situation wasn’t a simple misread; it stemmed from fundamental errors. The figures you shared—while we don’t need further debate—don’t reflect the true false positive rates which are certainly critical to be concerned about. A false positive indicates that we read a license plate, let’s say “FEDER,” and, presuming we are confident it’s “FEDER” reported as stolen in NCIC, we issue an alert.
It’s not merely about our accuracy but our confidence in it. This highlights the key aspect of technology: it will never reach absolute perfection, just as human eyesight won’t. The challenge lies in embedding redundant systems and robust limitations within the technology to reduce the potential for negative incidents.
In our specific example, say it’s “FEDER,” and we’re 99% confident for F, 99% for E, 99% for D, 99% for E, but only 80% for R—it wouldn’t trigger an alert due to lack of confidence.
Joel: Is this the current operational protocol?
Garrett: Yes, this is indeed how we operate. That’s a pivotal point, as we retain images and categorize them according to confidence levels. However, we set a high standard for notifying local law enforcement regarding NCIC alerts. Confidence needs to be ensured, even following comprehensive training on human verification processes or double-checking with NCIC for active warrants—mistakes are inevitable.
Recognizing that many—including you and I—find police stops unnerving, there’s a strong desire to achieve a 0% likelihood of wrongful traffic stops.
Joel: What is the necessary threshold? You mentioned 80% or 99%. What is it exactly?
Garrett: The specific figure is less critical; if I claim it’s set at 50%, you might ask how that is reliable. There’s a perspective here: it’s less than one in a million instances where an officer expresses uncertainty about an alert. That’s an extremely low count. If it were one in a billion, that would introduce another layer of trade-off.
From my viewpoint, I aspire to ensure that the OCR (Optical Character Recognition) performs optimally. I would argue we possess the leading license plate reading technology globally. Being the largest distributor likely correlates to superior capabilities. I maintain that we have the necessary safeguards to alleviate issues; your experience highlights constraints with NCIC, yet we’ve control over preventing wrongful stops due to OCR misreads, where we can make substantial improvements.
Joel: This smoothly transitions us into wider queries regarding the present situation and anticipated future. Let’s focus on your vision. As the founder and CEO of Flock, what’s your outlook moving forward?
As context, I reside in Plymouth, Minnesota, where there are currently 18 Flock cameras, although that doesn’t encompass every street corner. Other cities have more, others less. So, what’s your vision for the future? Will there be a Flock camera on every corner, in every school? Additionally, you are unveiling new technology beyond merely placing cameras atop streetlights. What does this future look like?
Garrett: First off, I appreciate the chance to share this vision as I’ve dedicated the last decade contemplating it. Like you, I’m a father of three and prioritizing their safety is paramount. They need to feel secure. After that, I aim to ensure they’re well-educated and grow into generous, productive members of society. If they don’t feel safe, everything else gradually disintegrates—and that holds true for many children. A prevailing lack of safety can destabilize societal foundations.
Looking at the broader view, we’ve historically analyzed crime and safety in two significant categories: emotional crime—crimes of passion—which have persisted for millennia. Unfortunately, we can’t put an end to these; they’ve existed as long as humanity itself. However, opportunistic crime is quite manageable. Approximately 95%-99% of criminal activities are opportunistic in nature. It’s exemplified by a thief walking by your parked car, believing they can steal it without consequences, and choosing to act on that impulse.
Places like San Francisco reflect this, where four years ago, committing crimes was surprisingly easy. A couple of years ago, Mayor Breed and Mayor Lurie approached us, stating, “Flock, we need support; criminal acts are rampant, our businesses are leaving, and residents are departing.”
We deployed hundreds of license plate readers along with our drones and real-time crime center software to optimize police efficiency. Now, it has transformed into one of the safest cities in the U.S. Mayor Lurie boasts one of the highest approval ratings of any major city’s mayor—he might be at the top currently.
In my vision for the future, I envision a landscape where cities mirror San Francisco, making committing crimes exceptionally challenging—so difficult that individuals decide against it altogether. While that vision might seem futuristic, it’s becoming a reality in San Francisco. I believe we can replicate that in Oakland, and currently operate in 7,000 cities. My ambition is to expand Flock’s reach to all 17,000 cities across America while fostering safety—this is within reach. We will continually innovate the technology necessary to achieve that goal while respecting societal norms where we operate, even if that sounds ambitious. I asserted as much a decade ago, and now it’s materializing at scale—it’s a profound moment of development.
Joel: Speaking of societal norms, you mentioned this topic. I want to clarify that we’re not venturing into political zones on the podcast nor doing so here. Yet, when it comes to societal norms—at one end we’ve individuals claiming, “If there are cameras on every corner, it essentially creates a dragnet surveillance of our lives without warrants, violating the Fourth Amendment.” Where do you personally stand with respect to these concerns about Flock today and where it might head tomorrow, especially regarding the warrantless surveillance aspect?
Garrett: Folks often have three main concerns or questions surrounding the company: privacy, constitutionality, and accountability.
In terms of the constitutional aspect, that’s relatively straightforward. Neither you nor I can unilaterally decide what constitutes a constitutional infringement. We rely on judges and elected officials to navigate that process, and we adhere strictly to the law and constitutional guidance. Thus far, it’s our perspective that the courts have affirmed this isn’t considered a warrantless search; it constitutes a valid product relevant to the Fourth Amendment. I don’t foresee any alterations there.
On the matter of privacy, we ask significant questions ourselves. Before Flock, ALPR data was indefinitely kept and sold for commercial usage by repo trucks to insurance companies, which I found quite troubling. Flock was among the first to set a 30-day default retention policy and made a commitment not to sell such data.
Here’s an intriguing statistic: last year, we assisted in solving over a million crimes utilizing Flock’s systems in coordination with local law enforcement. Now, hypothetically, if we reduced that retention period from 30 to 14 days, resulting in 30% fewer solved crimes—approximately 300,000 additional victims left without justice—would that be a trade-off anyone would readily accept?
You don’t need to answer right away—I wouldn’t put you on the spot—but my point is we shouldn’t determine that arbitrarily. It ought to be up to local officials, such as the Mayor of Plymouth and City Council, to make those determinations based on community values. Those elected representatives are accountable for making nuanced decisions within the established frameworks. Individuals can recommend reducing Flock’s functionality to cover just Amber Alerts or Silver Alerts, for example, which might result in zero data retention, focusing solely on alerts. My response would be that approach would net about 50% of Flock’s potential benefits. If that’s what you seek for Plymouth, we would be thrilled to facilitate that—however, you’d likely see a decrease in community safety. While it would be safer than before Flock was in place, it wouldn’t reach the levels experienced in San Francisco.
In a nutshell, we’re not advocating any particular stance as private citizens. I call Atlanta home, while you live in Plymouth. What we genuinely wish is that more elected officials would perform due diligence, engage the community, and establish reasonable standards that reflect their constituents’ values.
Joel: Let’s discuss the route of elected officials. I have seen multiple Congressional representatives—far more than just one or two—making posts on Instagram proposing bills to cut federal funding from municipalities that use Flock. I’m interested in your thoughts on that. To my knowledge, none have officially introduced these bills as of this recording, but discussions are certainly taking place.
Garrett: At a broad level, I would be thrilled to establish state-by-state regulations for this type of technology. That need is glaring because the speed of technological development has historically outpaced regulatory measures. That gap is typically manageable, but in this instance, authorities are lagging.
Proposing to defund approximately 80% of police department resources for utilizing technology to protect their citizens is incomprehensible. How is that a logical stance when there are far better alternatives for regulations?
I maintain that the probability of any individual actually endorsing that kind of reckless legislative action is minimal. The last time this was discussed—about six months ago—it barely got a footing before it was scrapped. I’d prefer to challenge those Congressional figures, rather than criticizing the dedicated personnel working to safeguard our communities; perhaps we should genuinely look into what necessary guidelines could be established to satisfy public concern.
It seems outrageous that currently, there is roughly a 50% chance of committing homicide without facing consequences—a reality we should find unacceptable. Denying law enforcement the tools that can elevate clearance rates and deter crime…
I bring this up because in Atlanta would be one notable case. Five years back, a woman was tragically murdered late at night in Piedmont Park—our largest park—and that case sadly remains unsolved. This is in part due to malfunctioning cameras. If those cameras had been functional, that family could have secured justice, but they can’t today, which I find appalling. We cannot accept such outcomes as a society, and I know we’re not comfortable with it in Atlanta.
Joel: We’ve covered Congress members and elected officials; I must get your insight on public backlash or towns opting out of contracts or refraining from renewals. I do want to clarify, I believe there’s been a considerable degree of misinformation and/or inadequate reporting regarding a few towns that claimed they “canceled” contracts; many were about contracts set to expire, which were simply not renewed, and this discrepancy frustrates me as a journalist. However, that’s beside the point.
I’m curious: what’s your perspective on public backlash—towns either discontinuing or failing to renew contracts? Flock isn’t a consumer product like iCloud where people opt-in; it’s something that’s imposed when driving on the roads, so I imagine that leads to some public pushback against the dragnet perception. We’ve talked about your view on that, yet I’m interested in your take on the public responses regarding discontinued contracts.
Garrett: It’s a fascinating development. In most businesses—for instance, taking your field as an example—when an advertiser or partner opts not to renew, it doesn’t typically grab headlines. Newly acquired advertisers don’t make headlines either. It’s interesting we’ve added roughly ten times as many new customers compared to those who have opted out in the last 90 days, but you won’t find that highlighted in the news because “This city adopted Flock” is a less compelling story than “This city halted their service.”
Additionally, it’s curious to note the media doesn’t highlight the numerous agencies that suspended services, saw crime rates rise, and then reverted back to Flock. That’s also part of the picture. I can’t say why this is so, but while some are stepping back, it’s their prerogative. What they’ll find, however, is a rise in safety risk, and eventually, they’ll reach out to us again, whether it’s in a week, a month, or a year. We’ve demonstrated value in improving public safety, and many cities that have experimented with this will discover the consequences of not using it.
Joel: Before we explore a rapid-fire section with whatever remaining time we have: you’re a private entity acting as the intermediary between the public and governmental bodies. You’re aiding law enforcement, but you aren’t a direct supplier. You offer a service, a SaaS, which is subscription-based—if clients stop paying, as we’ve discussed, the service cuts off; they lose access to the service. Yet your role is to serve the government while documenting the public’s vehicles or traffic. You aren’t government-run; this concept is rather different for this kind of service.
Garrett: This touches on an intriguing point and gives me an opportunity to clarify a few aspects. First off, and this has been misconstrued in various reports—hopefully, this helps in getting clear communication out—we do not possess the data; the municipalities retain control of the data. We do not access it. We merely act as the custodian, chosen for that reason based on customer feedback.
Questions arose from clients, asking, “Why can’t we just purchase equipment?” I would reply, “How’s that working for you? Are you camera experts regarding installation and upkeep?” The Atlanta case previously mentioned exemplifies issues arising when we overestimate government capabilities. Law enforcement excels at solving crimes, yet they aren’t precisely skilled in camera development and installations.
It’s essential to establish that our clients own the data. I believe one of the most challenging elements in developing this business stems from the last company in this field going public around 25 years ago. No one seems keen to engage with local government. This may explain the nature of our company, and I appreciate that local governments do require innovation and new technology for operational improvement despite the difficulty posed.
Joel: Josh mentioned something that aligns with what you previously noted, asserting that storage takes place on AWS (Amazon Web Services) and that data is segmented by municipalities which enables specific sharing capabilities among them.
Numerous incidents involving misconduct—misuse of your system by law enforcement—have made headlines; stalking women or repeatedly checking specific license plates, like under a hundred occurrences in a short window. It’s known that audit logs are crucial—a point Josh addressed as a method of tracking unethical behavior.
Do you incorporate multi-factor authentication within your system? The reason I ask: even my iCloud and Meta accounts utilize multi-factor authentication, streamlining the process of identifying individuals in audits for accountability.
Garrett: Yes, we do. In fact, we require multi-factor authentication, which represents a positive advancement since this policy has been in place for several years. Initially, it was optional. We recognized that in towns with minimal officers, it was unlikely they would provide both a government-issued laptop and mobile device—necessary for effective multi-factor authentication.
However, our perspective evolved, and we now mandate any version of MFA, asserting it’s superior to having none.
Regarding auditability and accountability, I see this as a broader challenge in law enforcement—we’re among the few systems implementing auditing and abnormal behavior detection effectively. The unusual behaviors displayed are flagged by our system—while not inherently illegal, they warrant further investigation. I wish more state regulations would require that every law enforcement system incorporate such sophisticated auditing with anomaly detection feature.
There’s progress to be made, and this is a key focus area. One additional detail, Joel: earlier today, I was speaking with an Ohio client who shared a success story involving ICAC (Internet Crimes Against Children). Thinking about the depths of societal miscreants—some of the worst individuals—he explained how a cold case was revived with Flock’s assistance and the perpetrator was apprehended. Incidents like this seldom make headlines, and there’s a societal tendency to shy away from such discussions.
I get the privilege of hearing these heroic stories from clients daily, where Flock assists in unearthing ‘the one needle in the haystack.’ I wish both the media and our clients could convey these narratives more effectively, understanding that while society prefers not to dwell on such dark topics, local and broader law enforcement are vital in apprehending those engaged in criminal activities.
While any exploitation is a significant issue that needs addressing and should hit zero, it’s imperative to recognize that last year, over a million crimes were resolved using this technology, showcasing that exploitation incidents remain exceedingly low compared to those numbers.
Joel: I’d forfeit my credibility if I don’t include this question even though it isn’t from my original outline; I know Flock frequently makes news both for good and bad reasons. The fact we’re convening here stems from my incident with law enforcement technology. I must ask about something that caught fire online: in a recent interview, you responded passionately to questions regarding individuals vandalizing or even shooting Flock cameras, characterizing them as “domestic terrorists.” Since then, you’ve publicly apologized, and I’ve seen the press coverage addressing that point. If we rewound time, how would you address that subject now?
Garrett: Clearly, I could have clarified the distinction between individuals committing felonies—where, to clarify, damage to government property constitutes a felony in numerous states. Although laws could change, currently, the statutes dictate that vandalism is illegal. I find the confusion surrounding individuals advocating for law-breaking perplexing. Since Flock was designed with the principle that people shouldn’t breach laws, I fail to understand how that perspective is considered controversial.
What I miscommunicated, as you pointed out, was the need to differentiate my remarks about lawbreakers from organizations like the ACLU or EFF, and now groups like DeFlock advocating for their viewpoints. I appreciate their efforts; even if we differ on specifics, I respect their right to express opinions. Much like I have a right as a business owner and father to speak out, individuals’ involvement in local politics is similarly vital. I’m unsure how involved you are in Plymouth, but I recognize your capacity to drive change in local matters; those issues often have a more direct impact on daily life than state or federal politics.
Joel: I truly value your time and thank you for setting this up. I’d love to continue this discussion at another point in the future, exploring aspects of technology, what your company can offer, and how it varies from competition—as there’s a wealth of topics to delve into.
Garrett: Fantastic! Thank you, and I look forward to our next chat.