I started the company nine years ago with a pretty simple mission, which is to live in a safer neighborhood. And that was our initial pitch. You know, if you look at our first two or three years in business, it was just neighborhoods. We sold a neighborhood, so we sold a license plate reading camera, which is pretty basic technology. A car drives by. We take a still image of vehicle. We read the license plate.

Langley explains the deceptively simple origin of Flock Safety, grounding a now-controversial surveillance network in a humble neighborhood-safety pitch.

North of a million crimes solved using Flock. North of 10,000 people found who were missing. And that's everything from our grandparents with dementia to our children. Like just last week in Atlanta, there was a stolen car that happened to have a 2-year-old in the back seat. And thankfully, because of very hard work by Atlanta Police Department and our cameras, that little kid was safely returned within the hour.

Concrete impact numbers anchor the safety side of the privacy-versus-safety debate with real human stakes.

What has been true for the last nine years um is we've always said this is this data is a liability to an asset. You should have a set retention period and that retention period for how long this data will be stored is going to determine the efficacy of the product. And so if you delete this data in a day or two, it's going to be less effective. It's still going to be better than nothing. And if you hold this for 30 days, it's going to be pretty effective.

Langley articulates the core tension in surveillance-system design: longer retention increases utility but also increases privacy risk and potential for abuse.

We think 90% of crimes will get solved within seven days of data retention. The 10% we're talking about or the 10% you read about in the news though. Oh, right. It's the homicide that there was never a 911 call because someone was killed and it takes five days to discover the body because of a welfare check and then it's a complicated case or it's look and I'm sorry to you know go to this topic. It's a rape victim who felt guilt, who felt shame, and she didn't call 911 until a week later because she wasn't ready.

The 90/10 data-retention split reveals that the hardest, most vulnerable cases — not routine theft — are the ones demanding longer retention windows.

We don't do facial recognition. We don't capture video. Um, we don't look inside the car. Um, we don't even allow in our system to search for people um for our license cameras. And that's in direct conflict with some of our biggest competitors who do all of the things I just mentioned.

Langley draws a clear product boundary that distinguishes Flock from more invasive competitors, but also reveals those capabilities exist in the market.

I I finally got to the bottom of it because I was like, why do you why do you hate this company so much? We sell a camera. Um and I realized this woman was like, Garrett, I I don't trust the police at all. And I was like, ah, that that's the root issue. And I was like, but why? And she was like, 'They lie, they cheat, they steal. I don't trust any of them.' And I'm like, 'Okay, man. What if you're what if you're right, I'm wrong.'

Langley admits that opposition to Flock is often not about the technology itself but about deep, legitimate distrust of law enforcement — a reframe that shifts the entire debate.

Anyone in a position of power, there's a higher likelihood of abuse. Look at CEOs, look at elected officials, look at police. When you're in power, abuse tends to follow not too far behind. And so when we met, you know, six or seven years ago, we had the audit log. And my assumption was like, this is good enough. This is what most police systems do is they have this audit log. But what I didn't realize is he was meeting with a large agency in Southern California and they've got something like a thousand officers a day that use Flock and imagine they're doing five or six searches a day. And imagine on once a month someone one person is supposed to review that log.

Langley candidly acknowledges that audit logs are nearly useless at scale without automation — a systemic failure he had previously dismissed as 'good enough.'

Here's a good example. Um, this is the easiest one. Let's say you're an officer and you'll say you're trying to stalk me. You would look for me. You'd say, 'I'm looking for Garrett's car. It's ABC124.' And you'd search for it and kind of figure out where it is. And then the next day, you'd search for me again. And then the next day, you'd search for me again. That's very abnormal. If you were truly trying to find me because I committed a crime, you would have put my tag on the hot list so that anyone else in the police department would have been also notified... If you're stalking me, you don't want anyone else to know what's happening.

Langley describes the behavioral signature of abuse — repeated single-officer searches without a shared alert — which became the foundation for Flock's automated abuse-detection tool.

We've caught a lot of bad cops and and a ton. Yeah. And it's it's a ton. It's it's more than I ever would hoped. Um, and you look at my home state here in Georgia where, you know, some of our customers have been with us for years. And I have to give those chiefs an applaud because they're not hiding behind it. They're publicly saying, 'I fired nine officers today, which is which is the the most severe offense in Georgia for for abusing the system. They're out of my police department. I don't want them.'

The scale of abuse uncovered by Flock's new automated tool — far exceeding Langley's expectations — validates critics' concerns and demonstrates that self-policing by departments was not working.

We now are requiring that as a mandatory. It's not even a default. It's a mandatory feature. You have to have turned on. We have to find these bad cops and get rid of them. They shouldn't have a badge. Amazing. And so this is very interesting, I think, because if we assume that, you know, let's just pick a number, 1% of cops do bad things. Some people might think it's all cops, some people might think it's 10%. I'm just gonna say 1%. It it's a very small percentage, but as a in a in a city with 30 or 40,000 police officers like New York City has, I believe, and I think Los Angeles 10 15,000 cops, 1% all of a sudden becomes 100 or 300.

Making abuse detection mandatory rather than optional is a meaningful policy shift, and the arithmetic on even a small misconduct rate reveals why passive oversight fails large departments.

My question to you is when I make a request, I'm a rank and file cop and I go do this. Why should rank and file cops be able to do this? Shouldn't there be like a higher level person who they put the request in? Or maybe there could be like a supervisor and that supervisor. So it says, I'm looking into this case. I explain why I'm looking into it. And then a senior person like a sergeant on duty. They have to say okay, you can run that search. And then a third person has to look at the sergeants okays and they put a note on it. You could also build that into the system which is you know I think like a double key system where two people have to say okay to this which now you're a dirty cop you got to find a dirty job.

Calacanis proposes a multi-party authorization model for database queries — a technical control that would structurally require collusion for abuse rather than just detecting it after the fact.

You are responsible now as a technology provider with the worst behaviors of your customers. And so the lesson there is even if you want to say hey we just sell a tool the reality of running this company since it is uh you know people have to give up some level of privacy to get that security you actually have to own it there's just no choice and I think previously your approach was hey it's a tool do with it what you will you made that choice which is not exactly an unreasonable thing to say people sell laptops and cameras and guns to a local police department they send tasers to a local police department. We don't say to the taser company, 'Hey, you're responsible when somebody gets tased.' And it was an inappropriate tase. It was an unjustified taser. But here we are. This is different. This is qualitatively different because America and Americans really value privacy.

Calacanis articulates why the 'we just sell a tool' defense fails for surveillance infrastructure in a way it doesn't for physical tools — a distinction with major implications for tech company liability.

Now I would prefer if every elected official understood how all this technology worked perfectly and made decisions that represented their community values. I don't think that's a reasonable expectation because that means they also need to know everything about data centers. They need to know everything about water works. like they're relying on their police chief to tell them what is best and that chief's relying on me to educate him on what I think is best. And so I do think the responsibility falls on us to give those strong defaults and then allow them to pivot against it.

Langley reveals a principal-agent chain where the vendor effectively shapes policy through defaults because elected officials rely on the police chief who relies on the vendor — concentrating real decision-making power at Flock.

We've been really consistent that um we never want to be a company that defines suspicion. Um that's really dangerous. Yeah. But for us I think there's it's really valuable that there's a human in the loop when it comes to public safety. We're talking about people's lives. We're talking about people's like it's it's really important stuff. So we always want to have a human in the loop.

Langley draws a firm line against predictive policing — refusing to let Flock define who is suspicious — while insisting on human-in-the-loop decision-making, both of which are contested norms in AI-assisted law enforcement.

You have to deploy this with third party proper attestation that it's that it's working as expected. It shouldn't be up to me. It shouldn't be up to my customers. Someone who actually doesn't want this to work should come in and say, 'Well, guess what? It it does actually like this is actually really good and it's making officers more effective. It's making officers more objective, less subjective. Like that should be our end goal is that AI is used in a way that increases the number of crimes they can solve and decreases the likelihood if they get the wrong person.'

Langley calls for adversarial third-party validation of AI claims — a governance standard rarely applied to law enforcement technology today.

The one of the first customers that chose to walk away did so because of facial recognition. They went online and googled ALPR facial recognition. They saw one of our competitors website and said, 'We're getting rid of this flock thing.' They didn't realize they were looking at the wrong person's website. And that's tough because like that's on me, right? That's on me to better communicate to to these decision makers that like here's what we do, here's what we don't do.

Cities are making consequential public-safety decisions based on misidentified product capabilities — a communication failure with real safety consequences on both sides of the debate.

I saw this one city in in Washington just just it's so sad. They turn the cameras off. The next weekend, a little boy gets shot, car drives off. It's gonna be a cold case. And luckily, the city next door still had Flock cameras, found the car, made the arrest, and look the that the kid that didn't live, which is horrible. At least their family knows justice was served.

The geographic patchwork of surveillance adoption creates a 'weakest link' effect where opting out of a network degrades safety for the opting-out community while others still benefit.

It's not spoken about as much, but it's worth looking at who is so against what we do and who is so for it. And it's typically people who don't have any other way to protect themselves, who need this, right? who want this. And and for those who have a gate and have a guard and have cameras and have a security system and they have Jaime's incredible doorbell camera, right, and they have all these products and these services, they're like, 'Why does why does anyone need to feel any safer?' It's like, 'Because most people don't have any of this stuff and they don't feel safe when they go to bed.'

Langley reframes the privacy-vs-safety debate as a class issue: opposition to public surveillance is a luxury position available mainly to those who can afford private security.

She was the victim of political violence. Um someone shot at her house and the police used a flock camera to find that person and lo and behold, to her credit, she changed. But she did change and she's now one of our largest advocates because she realizes you don't get to pick when you're the victim of violence. You don't get to pick. Sometimes it's random. Sometimes it's intentional, but everyone wants justice when something bad happens.

Personal victimization shifting a vocal opponent to an advocate illustrates how abstract privacy principles can collapse when confronted with concrete personal safety threats.

These drones live um in a kind of a HVAC control box, a dock, per se, on top of a roof, typically like a police precinct or a fire department. And when a 911 call happens, instead of sending an officer first, you can send a drone. And so there's this really interesting case not too far from you up in the Dallas Fort Worth area where a woman, you know, at late at night at a gas station called 911 and said, 'There's a man walking back and forth with a gun. I don't feel safe.' Um, which obviously it's not illegal to to garnish a weapon in the state of Texas, but she was scared.

Drone-first response reframes drones not as surveillance tools but as de-escalation tools — a use case that could dramatically reduce officer-involved shootings in ambiguous situations.

It's not a gun. It's a lighter. It's one of those toy like lighter things. That call the woman back and say, 'Hey, ma'am, it's just a lighter. You're okay.' And I share that story because so much of my businesses have related around solving crime, but it's more than crime. It's about safety. It's like, 'How do we help the community feel safe?' And this is an example where like we demitated, we mitigated what would have been a really bad situation for everyone.

A drone resolving a potential police shooting via visual verification before officer arrival is a concrete demonstration of how technology can prevent harm rather than just document it.

This is the next big like chapter for law enforcement is is doing a better job implementing drones than they've done with historical products because you look at something like body cameras and I'm not sure how much you follow this there was massive outrage on body cameras and now it's like wait a second this actually really good it's a better tool and and same with non-lethal weapons um and so no we we there's still more work to do there and to your point on the third party auditors you know one you've got a city manager that is effectively his or her job is to hold the police chief accountable and they can the audit logs. They can go in there themselves.

Langley draws a historical arc — body cameras were controversial, then accepted — to argue drones will follow the same pattern, implying public resistance is a predictable phase rather than a fundamental objection.

As an industry, you know, we're going to have to make these decisions sooner than later of what that default is because we have a default now, but there's very little state regulation. There's zero federal regulation outside of air safety, which is important, but but nothing yet on state regulation or city regulation for the drones.

Flock's drone program is operating in a near-complete regulatory vacuum — vendor defaults are effectively the only governance framework in place.