Latest / R: On Everything / Flock Safety: The Surveillance State Has a Logo

Transcript

Flock Safety: The Surveillance State Has a Logo · R: On Everything

• 0:24
  At the intersection of technology, publishing, leadership, and culture, I'm Ryan R. Campbell, and this is R on Everything. In this episode, we're doing a deep dive into flock safety and the burgeoning surveillance state. A little while back, I found a job posting that made me stop scrolling. After years in software engineering, I'd been looking for opportunities to bring more of my interests together: building things, after figuring out what's worth building, and helping people navigate change. The responsibilities lined up remarkably well with the work I'd done and, perhaps more importantly, with the kind of work I'd like to do next. Technology deployed at scale. Product strategy. Complex systems. Real-world consequences. Remote work. It was the kind of posting where you get several paragraphs in and begin unconsciously translating each bullet point into a story you could tell during an interview. There was only one problem: the job was at Flock Safety, and I couldn't do it. If you're unfamiliar with Flock, it describes itself as a public safety company. Its pitch is appealing: help find stolen cars, identify vehicles involved in shootings, bring missing people home. I suspect many of the people working there sincerely believe in that mission. Helping find a missing child would give someone good reason to feel proud of their work. That's part of what makes this worth talking about. Some of the most dangerous systems a society builds are created by smart, earnest people solving narrowly defined problems. How do we find stolen cars? How do we identify a vehicle involved in a shooting? Those are good questions. But the answer Flock and its customers have helped assemble is an infrastructure capable of recording the movements of enormous numbers of people who aren't suspected of anything. No matter how useful that infrastructure may be, I couldn't reconcile helping construct it with my own values. And usefulness is where this gets difficult. A camera can help solve a crime while also giving someone the means to track a person who hasn't committed one. To understand what we're being asked to accept, we need to look beyond the camera itself. You've probably seen a flock camera, even if you didn't know it at the time. The familiar versions are small black devices mounted on poles, often paired with rectangular solar panels and pointed toward roads, intersections, or neighborhood entrances. They're called automated license plate readers. You'll sometimes hear the abbreviation ALPR. But a plate number is only part of what they collect. Flock systems can also recognize a vehicle's make, color, body type, and other visible characteristics. Investigators can search for what the company calls a vehicle signature even without a complete plate number. Drive past a camera and, absent any suspicion that you've committed a crime, Your vehicle can become a timestamped observation. Pass another camera, and there's another observation. The power is in aggregation. One camera seeing your car at 8 in the morning provides a single piece of information. A network seeing it near your home, near your workplace, outside a medical clinic, and on the road to somebody else's house can begin to reveal patterns in your life. Where do you sleep? Where do you work? Which church, union hall, or protest do you visit, and how often? A car can have more than one driver, a sighting can be wrong, and passing a building doesn't prove you went inside. But the record can still give somebody a place to start looking, or an assumption to act on. Flock emphasizes that its license plate reader product doesn't include facial recognition. Okay, fair enough. But imagine I know where your car has been throughout the past week. Now tell me with a straight face that I know nothing about you. The scale is already enormous. An October statement from Senator Bernie Sanders's office put Flock's network at more than 120,000 cameras, producing around 20 billion vehicle scans a month. The same car can be recorded repeatedly, so that isn't 20 billion different vehicles. It's the repeated observations that make a movement history useful. Most of us produce those observations while doing perfectly ordinary things. Which is why the reassurance people reach for so often misses the problem. Why should I care? I'm not doing anything wrong. I understand why that feels reassuring. We're getting groceries, going to work, dropping the kids off at soccer practice. Why would anybody care? Well, somebody who wants to control you might care a great deal. In August, a Menasha police officer was sentenced to six months in jail for using flock cameras to track his ex-girlfriend. According to Wisconsin Public Radio. The person being watched didn't have to become interesting to a criminal investigation. She had to become interesting to somebody with access. I'm not doing anything wrong offers very little protection against that. There's also the question of what counts as wrong and who gets to decide. Attending a protest, visiting a clinic, organizing a union, worshipping somewhere your neighbors dislike— these can all attract unwanted attention without being crimes. Even when the people running an institution respect those freedoms, they won't be running it forever. Eventually, someone you don't trust will hold power. You don't even have to agree with your neighbor about whether that person is in office already to appreciate the need for limits. The Fourth Amendment doesn't say people shall be secure against unreasonable searches and seizures only when the government is run by the correct party. And catching an employee who breaks the rules only answers part of the problem. A search can comply with a department's policy and still reveal far more about innocent people than we're willing to let the government collect. An audit can flag a forbidden search. It can't settle whether the searches we permit are acceptable in the first place. That question gets harder when the department making the rules isn't the only department using the information. You might trust your police chief. You might know your sheriff personally. Maybe you believe they'd never use this system to interfere with somebody's medical care or political activity. But the value of a network comes partly from the other people connected to it. Permission to search information collected on your street can extend well beyond the officials you elected. In September, the Wisconsin Institute for Law and Liberty filed a lawsuit against Waukesha County and 11 municipalities over warrantless searches of the Flock database. The conservative law firm is asking for a warrant requirement, with valid exceptions. It isn't demanding that every camera come down. The complaint offers a striking example. During roughly a month in July and August, it says Muskego officers searched the database 315 times. Other agencies ran more than 900,000 searches that included Muskego's cameras during that same period. Those are allegations in a complaint, not findings by a court. The outside searches could include many camera networks at once. That number doesn't tell us how many Muskego residents were targeted or even whether a search returned a match. What it does describe is a system whose reach is much larger than the local department's own use. Residents need to know who else can search information collected on their streets, and under what conditions. If your town promises that the cameras won't be used for a particular purpose, how does that promise apply to somebody searching from another state? Officials need to explain how they can enforce local limits after sharing the information. Residents shouldn't have to learn the policies of distant agencies just to understand what their town agreed to. People have jobs. They have children. They shouldn't need to become surveillance researchers to follow a local purchase. Of course, the reason departments want access to a network this large is that it can help them find things. Any serious argument for limiting that access has to reckon with what might be lost. Green Bay Police Chief Chris Davis told Wisconsin Public Radio that the system had helped his department develop leads and solve cases. The department's first annual audit found no evidence of misuse. By the time of that August interview, it had also suspended sharing outside Brown County while reviewing its policies. The chief cautioned against simply crediting one piece of technology for a decline in crime. That's a distinction worth keeping in mind when we hear claims about how well these systems work. Here in Milwaukee County, the sheriff's office credited the cameras with contributing to a 43% decline in certain crimes around county parks between 2023 and 2026, according to the Wisconsin Examiner. That's the department's assessment. It doesn't establish how much of the decline came from the cameras, other policing, changes in park use, or broader trends. We should ask what's being counted and what the comparison shows. Otherwise, evidence that a camera helped in a particular case can become a much larger claim that the surveillance program caused crime to fall. But suppose we could establish that some crimes would be harder to solve without these systems. I think we should be willing to face that possibility directly. We place limits on investigations knowing those limits can make investigations harder. We require reasons, procedures, and independent review because a free society has interests beyond making the state's work as easy as possible. That can carry a real cost. A limit on surveillance might mean a lead isn't available. The right response is to explain why the limit matters, not pretend the cost disappears. The benefits also have to be weighed alongside harms that are harder to put in a recovery count: a person being tracked by an ex, somebody deciding against attending a meeting, or residents learning that their ordinary movements are producing information they can't control. Usefulness is not the only value that matters. This month, officials in Wisconsin have been making decisions that bring that trade-off out of the abstract. In one county, the people removing the cameras acknowledged that they'd helped solve cases. They took them down anyway. On October 7th, WXPR reported that Oneida County was ending its Flock partnership after 6 years. 6 county cameras had been taken down the previous week. Captain of Investigations Tyler Young said concerns had grown about data protection and who could access the information. He put the uncertainty plainly. I can't stand here and say with confidence that nefarious actors can't be accessing the data or where all that data is going. He wasn't announcing proof of a particular breach; he was acknowledging a limit on what his own agency could promise. The cameras had been useful, and the sheriff's office still concluded that continuing the partnership carried too much risk for people's privacy. Then there's Milwaukee County. On October 1st, the county board voted unanimously to remove flock cameras from county parks. County Executive David Crowley signed the measure on October 12th at Carver Park, near a camera covered with a black plastic bag. That action concerns county parkland. It doesn't remove every camera operated by every municipality in Milwaukee County, and it leaves room for replacement surveillance equipment. The distinction will matter when residents examine what gets proposed next. But the reversal itself tells us something: an installed camera and a signed contract don't make an arrangement permanent. There's a broader record of communities reconsidering, too. The Institute for Justice's October 16th update listed 277 local governments in its database of canceled license plate reader contracts. It excludes places that only paused operations or switched vendors. That's a count of documented cancellations, not cameras already removed. It also doesn't mean all those decisions happened recently— researchers can add older decisions as they find them. Even with those limits, there are now hundreds of examples of communities deciding these arrangements shouldn't continue. And lawmakers are proposing ways to make the limits extend beyond individual communities. Earlier this month, Bernie Sanders, Alexandria Ocasio-Cortez, and Jeff Merkley unveiled the Ban Flock Act, The House version was introduced on October 1st and referred to committees. The proposal would prohibit federal acquisition and use of license plate readers and their data, with specified exceptions. It would also restrict federal spending on the technology. State and local governments would have to adopt comparable prohibitions to receive assistance from several federal departments. The text has an exception for toll collection, with restrictions on how that information could be used. It also allows uses explicitly authorized by Congress under detailed safeguards. So the name doesn't mean a ban on every camera in every setting. And the proposal reaches license plate reader technology beyond the Flock brand. A separate, Republican-led proposal, the Flock Off Act, would restrict federal funding for license plate readers and biometric surveillance cameras. It also had a Democratic original co-sponsor. Those proposals differ. Agreement that something needs limiting doesn't settle where the line belongs. It does give people who disagree about much of American politics a concrete place to work together. It also forces a question that can disappear when the debate centers on one company: if the same power is available from another vendor, what exactly are we trying to stop? Flock has become the recognizable brand. So getting rid of the brand can begin to look like getting rid of the problem. We already have a Wisconsin example. In July, UW-Madison Police announced that its Flock contract had expired and wouldn't be renewed. In the same announcement, it said it still supported license plate reader technology and was exploring other vendors. The department was clear about its intentions. Vehicle surveillance remained a capability it wanted to purchase. And on October 16th, CT Insider reported that Fairfield, Connecticut had replaced Flock with 54 Axon license plate readers. The police chief described tighter controls, including restricted access, case numbers, and monthly audits. Those controls could make a difference, and we should examine them on their merits, but the cars are still being recorded. The goal isn't to deflock America only to re-Axon it. The goal is to decide what powers we're willing to grant the state. That means asking what a replacement collects, how long the information remains available, who can search it, and what legal authority they need. A different logo doesn't answer those questions. Nor does an announcement about new safeguards tell us which protections apply to the system down the street. In August, Flock announced a 7-day default retention period. Existing customers, however, would keep their current periods unless they adopted the change. The company also said case codes and its audit assistance feature, which flags unusual activity for review, would become mandatory for law enforcement customers by the end of the year. Those are commitments with conditions and a rollout schedule. We shouldn't treat them as proof that every customer already has the same safeguards. And even if every customer adopted the shorter retention period, a week is long enough to include a medical appointment, a few nights at somebody else's house, or a meeting you wanted to attend privately. Better controls can reduce some risks. The decision to collect everyone's vehicle sightings still needs its own justification, from the people building the system as well as the people buying it. This is where I keep coming back to that job posting. There are interesting technical problems to solve in a system like Phlox. How do you make a search faster? How do you improve a vehicle match? How do you help an investigator find the relevant observation among millions of irrelevant ones? If you spend your days working on those problems, doing the job well can begin to feel like the whole responsibility. Your team releases an improvement, and a customer sends a grateful email about what it helped them accomplish. Meanwhile, the people whose movements make the system useful remain outside the conversation. That's why I couldn't separate the work from the power it creates. An engineer doesn't set every department's policy, but the features we build help determine what those departments can do. Making an invasive search effortless is a consequential decision even when somebody else gets to press the button. The people purchasing the system have a responsibility here too. A department might begin with a genuinely urgent problem and a limited budget. The camera looks affordable, the neighboring department likes it, and the vendor can point to a case in which it helped. It's easy to see how the purchase happens. What's harder to see in that meeting is the cumulative result of many other meetings just like it. Another neighborhood connects. Another agency is granted access. A new search feature becomes available to customers who already have accounts. The surveillance state doesn't need soldiers installing cameras in broad daylight. A thousand reasonable procurement decisions will do. Those decisions deserve a public conversation proportionate to the power they create, before the system becomes something everyone is expected to live around. We make surveillance politically expensive by bringing that conversation to the officials who can change the decision. Start with the cameras in your community. Deflock is a volunteer project that maps license plate readers from multiple vendors. Use it to look along your commute or around your neighborhood. Then, find out who owns the cameras, who approved the purchase, and when the contract comes up for renewal. Then, follow the information. Which other agencies can search it? Could an outside agency make a request through a local officer? How long is it kept? Does a historical search require a warrant, and who checks whether a stated reason is legitimate? Ask for the policy and sharing arrangements in writing, use public records requests, and support the reporters and local groups already doing that work. The Wisconsin stories we've been discussing exist because people kept asking questions a sales presentation wasn't going to answer. At a council or county board meeting, ask officials what evidence would cause them to end the program. If every crime decline proves the cameras work, every crime increase proves they need more cameras, and every misuse proves they need better training, we've designed a program that can never fail its own evaluation. We can demand public hearings before purchases and renewals. We can push for warrants for historical location searches, short retention periods, independent auditing, and enforceable limits on sensitive tracking. And we can argue that some capabilities shouldn't be purchased at all. That option has to remain available. Otherwise, officials are asking the community how to carry out a decision it was never allowed to reject. I understand why people get angry enough to want to fight the cameras themselves. But destroying a device doesn't destroy the policy that put it there. A camera can be replaced, a pole repaired, a vendor changed. The more durable work is changing what officials are authorized to buy and what anyone is authorized to do with the information. I don't expect everybody to arrive at the same answer about every use of a license plate reader. I do expect the people who want to deploy them to explain why everyone else should accept the consequences. Residents shouldn't have to prove that an abuse will happen to them before their objection counts. Freedom sometimes requires accepting that dangerous people will possess freedoms too, and privacy sometimes requires accepting that investigators will will have to investigate. Rights that exist only when their exercise is convenient for the government aren't rights. The state should not know where everyone is simply because technology has made knowing possible. The cameras are already here. Their continued presence is still a decision someone has to make. Make them make it in public. The officials approving these contracts answer to us. Act like it. Thanks for listening. For more content like this, you can subscribe to the publication and learn more about Our On Everything at ouroneverything.com. Until next time, I'm Ryan R. Campbell. Onward, upward, and keep curious. Audio assets for Our On Everything are made in collaboration with ElevenLabs. For more on the Our On Everything AI content policy, visit ouroneverything.com.