• The stated purpose and the actual network
• How the cliff works
• Flock’s response proved the pattern
• This has happened before
• The data breach that nobody needed
• Where this goes
  • Sources

On January 13, 2026, the Santa Cruz City Council voted 6-1 to terminate its contract with Flock Safety. Five weeks later, Mountain View voted unanimously to do the same after an audit found the ATF, Air Force, and GSA had all been accessing their camera data without authorization. Between those two votes, Flagstaff turned off all 32 of its Flock cameras overnight, Staunton pulled its cameras down within three weeks, and Denver’s mayor—who had twice bypassed his own city council to keep the cameras running—announced he was ending the contract too.

Forty-six cities have now rejected Flock Safety or similar license plate surveillance systems. In at least five states, people are physically destroying the cameras themselves. A 41-year-old man in Suffolk, Virginia named Jefferey Sovern methodically dismantled 13 of them over six months using vice grips. Someone in Oregon spray-painted one and left a note reading “Hahaha get wrecked ya surveilling fucks.” In La Mesa, California, two cameras were found smashed on a median.

I’ve been trying to understand what changed. Not why people dislike surveillance—that’s old news. But why the same cameras that were tolerated for years are now being pulled off poles by city councils and citizens alike. What I keep coming back to is a pattern that seems to repeat: surveillance systems aren’t destroyed because they watch people. They’re destroyed when people discover the watching was for something other than what they were told.

I’ve been thinking of this as the Compliance Cliff. Trust doesn’t erode here. It falls off a ledge. The cliff isn’t triggered by surveillance existing. It’s triggered by surveillance lying.

§The stated purpose and the actual network

Flock Safety’s pitch to police departments is simple: automated license plate readers that help catch stolen cars, find missing people, and solve crimes. The company operates roughly 80,000 cameras across 49 states, performing over 20 billion vehicle scans per month. Their valuation hit $7.5 billion in March 2025. The technology is real and it works. Nobody disputes that.

What triggered the cliff was a series of revelations about what else the cameras were doing.

In May 2025, the Electronic Frontier Foundation reported that a detective in Johnson County, Texas searched Flock’s camera network with the stated reason: “had an abortion, search for female.” The search covered 6,809 different Flock camera networks and more than 83,000 cameras nationwide, including in states like Illinois and Washington that have laws specifically protecting people seeking abortions. When reporters asked about it, both the sheriff’s office and Flock initially misrepresented the search as a “missing person” case. The EFF’s October 2025 follow-up investigation, working from a sworn affidavit by the lead detective, showed it was a death investigation of a “non-viable fetus” in which prosecutors were consulted about potentially charging the woman.

A University of Washington Center for Human Rights report from October 2025 found that at least eight Washington state law enforcement agencies had enabled direct sharing of their Flock camera networks with U.S. Border Patrol. Audit logs revealed apparent “back door” access by Border Patrol to the networks of at least ten additional police departments that had not explicitly authorized those searches. A separate 404 Media investigation found more than 4,000 nationwide and statewide lookups by local police that were done either at the behest of federal immigration agencies or with an immigration focus.

Then came the school cameras. An investigation by The 74, an education news outlet, found that over one month in late 2025, more than 3,100 police agencies conducted more than 733,000 searches on cameras at the Alvin Independent School District south of Houston, a district with 30,000 students and eight Flock cameras. Immigration-related reasons were cited 620 times by 30 law enforcement agencies across four states. Civil immigration searches outnumbered criminal immigration searches by more than two to one.

To be direct about what that means: cameras installed to protect school kids were being used to track their parents.

§How the cliff works

When surveillance does what it claims to do, most people accept it. This is true even when they don’t love it. Speed cameras are annoying. Security cameras in parking garages are fine. License plate readers that catch stolen cars are popular enough that thousands of communities signed up willingly.

The cliff happens when people discover a gap between stated purpose and actual use. A specific, documented gap that can be described in a single sentence: “The cameras that were supposed to catch car thieves were tracking abortion patients.” “The cameras at the school were being searched by immigration agents.” “The security cameras had a backdoor that let the Air Force in.”

Mountain View’s Police Chief Mike Canfield demonstrated this precisely. After discovering that his department’s 30 cameras had been accessed by the ATF, Air Force, and GSA without authorization—and that Flock had set one camera to “nationwide” access without permission—he said he “no longer has confidence in the Flock system.” Not that he’d lost some confidence. That he no longer had it.

That’s the cliff. Confidence doesn’t erode gradually under these conditions. It falls off a ledge. And the reason is structural. When a surveillance company says “we help catch car thieves” and the data shows they also helped track immigrants and abortion patients, the problem isn’t a policy tweak. The entire justification for public acceptance is invalidated. You can’t return to baseline trust by promising to stop doing the bad thing, because the existence of the bad thing proves the promises weren’t reliable.

§Flock’s response proved the pattern

In December 2025, Flock CEO Garrett Langley sent an email to law enforcement clients that described the company and its partners as being under “coordinated attack” from activist groups that “want to defund the police, weaken public safety, and normalize lawlessness.”

The Staunton, Virginia Police Chief Jim Williams, who received this unsolicited email, responded by canceling the city’s contract. His reply to Langley: “What we are seeing here is a group of local citizens who are raising concerns. These citizens have been exercising their rights. In short, it is democracy in action.”

The Langley email accelerated the cliff rather than slowing it. It revealed that Flock’s internal narrative didn’t match the external reality. Residents weren’t asking for lawlessness. They were asking why the cameras they’d agreed to for public safety were being used for federal immigration enforcement. Langley’s response told them the company didn’t understand the question.

Evanston, Illinois captured this dynamic in physical form. The city ended its Flock contract, and Flock reinstalled the cameras anyway. Evanston then covered them in plastic sheeting to prevent them from photographing license plates. If you want a visual metaphor for the Compliance Cliff, it’s a plastic bag over a surveillance camera because the institution that put it there stopped believing it was doing what it said it was doing.

§This has happened before

The pattern isn’t new. What’s new is the evidence. We used to argue about whether surveillance was creepy. Now we have audit logs showing exactly who searched what and why.

In 2019, San Francisco became the first U.S. city to ban facial recognition technology. By 2020, eighteen cities had similar bans. Then crime rates rose. New Orleans reversed its ban. Virginia reversed its ban. San Francisco police were caught violating theirs at least thirteen times—six acknowledged, seven more uncovered by reporters who found officers asking neighboring agencies to run facial recognition searches on their behalf.

The bans didn’t hold because the compliance cliff works in both directions. When surveillance is removed and crime goes up, the public crosses a different cliff: they discover that the absence of surveillance has costs they hadn’t factored in. New Orleans reversed its ban specifically because of a rising murder rate. The mechanism is the same. The gap between what people were told (“the ban will protect your privacy with no downside”) and what happened (“murders kept climbing”) triggered its own collapse of trust.

Internationally, the pattern is even more striking. In Italy, a figure known as “Fleximan” destroyed 15 speed cameras across the Veneto, Piedmont, and Lombardy regions in 2024 using an angle grinder, becoming a folk hero with street art in his likeness. Italy has 11,300 speed cameras, the most in the European Union. The French Yellow Vest protesters destroyed 60% of France’s speed cameras. London’s ULEZ cameras were systematically targeted.

But here’s the thing about the European camera destruction: it was about money. People saw speed cameras as revenue extraction disguised as safety. The stated purpose was “slow down to save lives.” The perceived actual purpose was “generate fines to fund the government.” Whether that perception was accurate matters less than the fact that a gap opened between stated and perceived purpose, and the cliff followed.

The Flock situation is that same mechanism applied to something more consequential than speeding fines. The stated purpose was crime prevention. The actual use included tracking immigrants, finding abortion patients, and letting the Air Force browse license plate data from a mid-size California city. The gap between stated and actual use is wider than anything Fleximan was reacting to. And unlike in Italy, the evidence isn’t a perception. It’s in the audit logs.

§The data breach that nobody needed

Then there was the data breach, which somehow made everything worse. In late 2025, at least 60 of the company’s Condor cameras were found streaming live to the internet without authentication. A YouTuber named Benn Jordan discovered it. Reporters from 404 Media confirmed it by accessing the feeds, watching about a month of archival footage, and even manipulating the cameras to track their own movements.

Flock called it “an isolated configuration issue and not indicative of a broader or ongoing concern.” Mountain View’s subsequent audit also found that over 250 California law enforcement agencies had searched the city’s camera data without authorization for over a year.

In the context of the Compliance Cliff, the breach isn’t a separate problem. It’s an accelerant. It shows that the surveillance infrastructure isn’t just doing things it shouldn’t be doing. It’s failing to secure the data it’s collecting while doing them. The contract with the public was: give us your license plate data and we’ll catch criminals. The reality was: we’ll give that data to immigration agents, military agencies, and accidentally to the open internet.

§Where this goes

The Compliance Cliff isn’t about whether surveillance is good or bad. It’s about the gap between what a surveillance system says it does and what it actually does. Wherever that gap opens wide enough for a regular person to describe it in one sentence, compliance collapses.

This means the cliff is predictable. You can look at any surveillance system and ask: is this doing only what it said it would do? If the answer is yes, the system will probably be tolerated. If the answer is “yes, plus these other things we didn’t mention,” the clock is running.

Colorado’s new bipartisan legislation—the Protecting Everyone from Excessive Police Surveillance Act, which would require warrants for ALPR data access and block sharing with federal immigration authorities—is one response. DeFlock, an open-source project that has mapped over 16,000 camera locations worldwide, is another. The guy in Virginia with vice grips is a third. They’re different actions driven by the same trigger: the discovery that the cameras were doing something they weren’t supposed to be doing.

Flock has started implementing search filters that block access to Illinois license plate data for abortion or immigration reasons. Denver switched to a different vendor. These are attempts to close the gap. Whether they work depends on whether the gap stays closed, which depends on whether the surveillance industry can resist the structural incentive to open it again.

The incentive to open it is strong. Federal agencies want access. The data is sitting there. The side-door mechanism—where federal agents ask local cops to run searches on their behalf—doesn’t require Flock’s cooperation or knowledge. One detective with a login can search 83,000 cameras for any reason they type into a text field.

I don’t know whether the cliff holds. What I know is that forty-six cities have already gone over it, and cameras are being pulled down faster than Flock can put them up. The system that was supposed to watch everyone is now being watched itself, and the people doing the watching have vice grips.

§Sources

• “Americans are destroying Flock surveillance cameras,” TechCrunch (Zack Whittaker, Feb 2026)
• “Across the US, people are dismantling and destroying Flock surveillance cameras,” Blood in the Machine (Feb 2026)
• “She got an abortion. So a Texas cop used 83,000 cameras to track her down,” EFF (May 2025)
• “Flock Safety and a Texas sheriff claimed a license plate search was for a missing person. It wasn’t,” EFF (Oct 2025)
• “Leaving the Door Wide Open: Flock Surveillance Systems Expose Washington Data to Immigration Enforcement,” UW Center for Human Rights (Oct 2025)
• “ICE Taps into School Security Cameras,” The 74 (Dec 2025)
• “ICE taps into nationwide AI-enabled camera network,” 404 Media (2025)
• “Mass surveillance fears push Silicon Valley city to scrap automated license plate readers,” Mercury News (Feb 2026)
• “Flagstaff cancels controversial contract for Flock Safety cameras,” AZ Family (Dec 2025)
• “Denver ends Flock contract,” Denverite (Feb 2026)
• “‘Democracy in action’: Cities react to privacy concerns by canceling Flock surveillance contracts,” SAN (Feb 2026)
• “Evanston, Oak Park end contracts with Flock Safety,” ABC7 Chicago (2026)
• “Flock CEO goes ballistic on critics,” ACLU (2026)
• “Big Brother left the door open: Flock’s AI surveillance cameras exposed to the internet,” PetaPixel (Dec 2025)
• “EFF’s Investigations Expose Flock Safety’s Surveillance Abuses: 2025 in Review,” EFF (Dec 2025)
• “Flock police cameras scan billions per month, sparking protests,” NBC News (2026)
• DeFlock.org (46 cities rejecting Flock/ALPR systems)
• Colorado PEEPS Act (Senate Bill 70), passed Senate Judiciary Committee 5-2 (Feb 2026)

Originally published at https://noahaust2.github.io/strategist-dashboard/blog/the-compliance-cliff.html