Breaking the “Safe” Illusion: Why AI-Powered Police Cameras Threaten Your Privacy
By: Casey Cannady : technologist, traveler & unapologetic privacy hawk
TL;DR
All across the U.S., cameras you barely notice are logging license plates and building a record of where everyday people drive. Many are owned and run by venture-backed companies like Flock Safety, not by your police department. When the AI gets a plate wrong, innocent people end up face down on the pavement at gunpoint. “Hot lists” alert officers without a judge ever signing anything. And commercial plate databases sell sightings to lenders and insurers. Benn Jordan shows that adversarial patterns can confuse these cameras, but citizens should not need hacks to keep basic privacy. The fix has to be law, transparency, and regulation.
The video that sparked this post: Benn Jordan, “Breaking The Creepy AI in Police Cameras”
The Rise of Invisible Surveillance
These aren't just traffic cams anymore. They're automated license plate readers (ALPRs): high-speed cameras that capture every plate that passes, along with the location, date, and time, and upload it to a central server. Taken together, those records paint an intimate portrait of a driver's life. And increasingly, the cameras are owned by private companies like Flock Safety, not by your local police department.
That means the images and data (where you shop, where you drove, even what's on your bumper) end up in a network controlled by corporations, not courts.
Put simply: your daily commute has become an entry in someone else's database.
When AI Gets It Wrong, People Pay the Price
These companies market public safety. The reality is messier, and sometimes dangerous.
- Aurora, Colorado, August 2020. Police stopped Brittney Gilliam's SUV because its plate matched a stolen motorcycle, registered in a different state. Officers held her family at gunpoint and put children as young as six face down on the pavement. The city settled for $1.9 million in 2024.
- San Francisco, 2009. An ALPR misread one digit of Denise Green's plate. Officers never visually confirmed it, and she was pulled from her car at gunpoint, handcuffed, and forced to her knees. The Ninth Circuit later revived her lawsuit.
- Taxpayers foot the bill when predictable “AI errors” turn into multi-million-dollar settlements.
This isn't hypothetical “what if” fearmongering. It's happening now.
A Business Model Built on Your Data
Flock Safety's playbook looks less like community safety and more like Silicon Valley venture capitalism:
- Lease, don't sell: police departments, and sometimes HOAs and retailers, rent cameras on subscription.
- A network, not a camera: every new customer makes the shared network more valuable to every other customer.
- Venture-backed scaling: in March 2025, Flock raised $275 million at a $7.5 billion valuation.
If it feels like the surveillance tail wagging the public safety dog... it is.
From Hot Lists to Human Rights Problems
One of the most troubling features is the “hot list.” A car gets flagged, and every time a camera sees it, police get a notification. No judge. No warrant.
In United States v. Jones (2012), the Supreme Court held that attaching a GPS tracker to a car and following it for weeks is a Fourth Amendment search. A dense enough camera network can reconstruct much of the same movement history without anyone ever touching the car, and without the constitutional question ever being asked.
This isn't just about traffic tickets. It opens the door to misuse: stalking an ex, tracking political opponents, or enforcing private morality.
Big Data Makes Bigger Problems
Police cameras are only half of it. Readers on tow trucks, parking lots, and mall security vehicles have been feeding private plate databases since at least 2013, and companies like Digital Recognition Network openly market that location data to lenders and insurers. Combine ALPR sightings with retail logs, social media breadcrumbs, and your car's own telematics, and insurers, banks, and government agencies get a 360-degree view of your life.
That kind of profiling can:
- Raise your insurance premiums
- Affect your loan eligibility
- Decide whether you get targeted in an immigration raid
And the scariest part? You never opted in.
Can We Fight Back? “Confusing the AI”
In the video above, Benn Jordan demonstrates adversarial patterns: subtle visual designs that humans barely notice but that can make cameras misread or ignore a license plate. Think of it as digital camouflage for the AI's eyes.
The takeaway: it is possible to resist automated tracking. But it is likely not legal for everyday use, because obscuring a license plate is itself illegal in many places.
The bigger point Jordan makes, and I echo, is that citizens shouldn't need hacks and workarounds to protect basic privacy. The solution has to come from law, transparency, and regulation, not DIY evasion.
Why We Need Stronger Protections
The European Union already enforces the GDPR, which sets clear expectations about data collection, purpose, and consent. The U.S. has no federal equivalent, while venture-backed surveillance rolls out neighborhood by neighborhood. If we don't act soon:
- Surveillance creep becomes normal.
- Everyday privacy becomes a luxury good.
- Tech errors keep eroding trust in both AI and law enforcement.
Privacy is not a perk. It's a human right.
What You Can Do Today
- Stay informed. Don't ignore those small black boxes perched above intersections and store entrances.
- Ask questions. If your HOA, city, or store installs ALPRs, demand answers: who owns the data, how long is it kept, and who can search it?
- Support legislation. Push for GDPR-like protections in U.S. law, with short retention limits for ALPR data.
- Protect your digital footprint. Limit what you share online. “Free” apps and retailers are often bigger trackers than the government.
- Build awareness. Talk with friends, clients, and communities about the risks. Awareness is the first firewall.
Final thought: technology itself isn't the enemy. Opaque, profit-driven deployment is. As a technologist, I know we can build powerful tools that balance security with rights. But without transparency, oversight, and genuine consent, these cameras aren't safety devices. They're silent stalkers. And we deserve better than living under their gaze.
Sources & Further Reading
- Benn Jordan: “Breaking The Creepy AI in Police Cameras”, the source for the adversarial pattern research.
- EFF Street Level Surveillance: Automated License Plate Readers, how the technology works and how it has been abused.
- NBC News: Aurora family held at gunpoint receives $1.9 million settlement February 2024.
- ACLU: “San Francisco Woman Pulled Out of Car at Gunpoint Because of License Plate Reader Error”, and EFF: on the Ninth Circuit ruling.
- TechCrunch: Flock Safety raises $275M at $7.5B valuation March 13, 2025.
- NBC News: “License plate data not just for cops: Private companies are tracking your car”.
- Oyez: United States v. Jones, and GDPR.eu: What is GDPR?
Sourcing note: incidents, funding, and case law are linked to their reporting or primary sources. The argument about business models and what should change is my opinion. Revised September 2026: two dead links and one unresolvable link were replaced, an unsourced anecdote was replaced with a documented case, and the Flock funding figure and the Supreme Court reference were made precise.
Connect with Casey
Have a story, a question, or want Casey to write about a specific topic? DM me and tell me which story you want next.
| Websites | |
| Threads | |
| Bluesky | |
| YouTube |
Casey writes about privacy, surveillance, cybersecurity, nomadic life, and navigating the world as a late-diagnosed AuDHD adult. New posts drop on my professional website.