You are driving home on a normal day. You are not speeding. Then red and blue lights fill your mirror. The officer says you failed to signal, or that your window tint is too dark. You think it is bad luck.
Maybe it was. But a growing number of reports describe AI police traffic stops, where software flags a car as suspicious and an officer finds a small reason to pull it over. The traffic violation is the excuse. The real reason sits in a database you cannot see.
By the end of this post you will know how the system works, what the documented cases show, which fears are overblown, and what is worth doing about it.
What actually happened
The story starts with cameras. Automated license plate readers, or ALPRs, are devices that capture a vehicle’s plate and other identifying features. They upload that data to a massive, searchable database [4]. Think of a camera on a pole that never forgets a plate and never stops looking.
Those cameras are not new. What is newer is the software on top of them. According to the Associated Press, as quoted in a coalition letter, an algorithm flags vehicles as suspicious based on where they came from, where they were going and which route they took [4].
Flock Safety is one of the biggest names here. The company describes its offering as the largest public-private safety network [3]. Its system lets police search the nationwide movement records of any vehicle that comes to their attention [2]. A feature called Multi-State Insights alerts law enforcement when suspect vehicles have been detected in multiple states [2].
Federal agencies feed into this too. Customs and Border Protection uses data from private plate reader companies, including Flock Safety, Rekor and Vigilant Solutions [1]. Border Patrol uses plate readers disguised as traffic barrels and flatbed trailers [4]. Researchers have identified more than 40 suspected covert readers along roads in San Diego and Imperial counties in California [4].
Once a car is flagged, the pattern is simple. Local law enforcement is told about drivers on “abnormal” routes, then makes a pretextual stop [1]. A pretextual stop is a traffic stop made for a minor reason, when the officer really wants to look into something else.
The stated reasons are ordinary. Drivers are stopped for speeding, failure to signal, the wrong window tint or even a dangling air freshener blocking the view [4]. The true reason behind the stop is often kept secret, even from courts [1].
The Alek Schott case
One case shows how this can go. Alek Schott was stopped in Texas after a deputy said he drifted out of his lane. Dashcam footage later confirmed that Schott never drifted. Even so, he was questioned for 10 minutes before the deputy called for a drug dog [1].
The sheriff who pulled him over had received information about the driver from a WhatsApp group called Northwest Highway. State and federal officers used it to trade tips on allegedly suspicious vehicles [1]. So the lane violation was not the real reason for the stop.
Officers frequently leave empty-handed, as they did with Schott [1]. The Institute for Justice sued on his behalf. The suit says Bexar County, Texas, and the officers who stopped him violated his Fourth Amendment right to be free from unreasonable searches and seizures [1].
Is this new?
Not entirely. Pretext stops have a long history. Operation Pipeline is a DEA program that trains police officers across the country on drug interdiction on roads [5]. In 1999 the ACLU wrote that the program was “instrumental” in increasing the use of pretext stops by law enforcement [5].
So the habit of stopping someone for one reason while looking for another is old. What changed is the source of the hunch. Before, it was an officer’s gut feeling on the roadside. Now it can be a score or alert built from months of camera records across many states.
The verdict
This is bad for privacy. Here is why.
Our position is that surveillance should be minimal and just enough to protect people. A system that logs where ordinary cars travel, then flags the ones that look odd, is far past minimal. It treats everyone’s movements as evidence waiting to be read.
The weak spot is the lack of public knowledge. The ACLU’s Jay Stanley says that because Flock is a private company, we know nothing about the nature of the algorithm or algorithms it uses [2]. If you cannot see how a driver gets flagged, you cannot check whether the rules make sense. A court may not see them either, since the real reason for a stop is often kept secret [1].
Now the other side. Police do have reasons to want better tools. Catching people who move drugs or run criminal networks is a legitimate goal. The problem is not that police want leads. The problem is that the leads come from a hidden system, and the person stopped gets no way to challenge it.
There is also a fair question about how big this is. The research here does not tell us how many stops happen this way. It gives us documented cases, a described pattern and a set of tools. Nobody outside the system seems to have a full count. So “police are pulling everyone over” would be wrong. “Police have a powerful, secret way to pick who gets pulled over, and it is spreading” is what the evidence supports.
Why it matters
Take Jake, who drives a long highway for work every week. A pattern that looks normal to him, the same route at the same hours, could look odd to a program that knows nothing about his job. A program has no way to ask him.
Then there is Sarah, who drives to visit family in another state a few times a year. Multi-State Insights alerts police when suspect vehicles show up in multiple states [2]. Sarah crossing state lines is not suspicious. But a tool built to spot cars that cross state lines does not know that.
Most coverage skips one point. For the driver, a stop that ends with “have a nice day” still costs something. In Schott’s case, that meant 10 minutes of questioning and a drug dog, with no violation behind it [1]. The harm is not only what officers find. It is being held and searched without a real reason.
Where you live also matters. California law prevents state and local agencies from sharing plate reader data with out-of-state agencies, including federal agencies involved in immigration enforcement [4]. Laws like that show the data flow can be limited. Most places have no such rule that this research covers.
What you should do
You cannot hide a license plate. It is on the outside of your car, by law. A VPN does nothing here, because the camera reads the car and not your phone or your internet connection. Do not spend money on gadgets that promise to fix this.
Here is what is worth doing, in order of urgency.
- Stay calm if you are stopped. Be polite. The roadside is the wrong place to argue about why you were stopped.
- Write down what happened. As soon as you can, note the time, place, the reason given, how long it lasted and whether a dog or a search came up. Dashcam footage helped in Schott’s case [1], so a dashcam of your own can be worth having.
- Talk to a lawyer if you think the stop was wrong. A blog post cannot tell you your rights in your state. Groups like the Institute for Justice have brought cases like Schott’s [1].
- Look up your local rules. Ask your city or county whether it uses plate readers and who can see the data. This is a short search, and local meetings are where these contracts get approved.
You can skip the rest. Do not change your route to look “normal.” Do not stop visiting family or driving at night. The odds that any one driver is stopped this way look low from what we know, and living around a camera system is not a reason to stop living.
Common mix-ups
“A flagged car means the driver did something wrong.” No. The pattern flagged is a route or travel history, not a crime. Officers often leave empty-handed [1].
“The stop reason on the ticket is the real reason.” Not always. The true reason is often kept secret, even from courts [1].
“This only happens near the border.” Border Patrol readers are one part of it [4]. But private networks like Flock’s cover far more ground, and CBP draws on them [1][2].
“Cameras are only used to find stolen cars.” The tools described here also search where any car has been and flag trips that cross states [2].
For the technically curious
The reporting does not describe the scoring method. The ACLU says the public knows nothing about the algorithms Flock uses [2]. What is documented is the input: the origin, destination and route of a vehicle [4]. The inputs come from plate reads uploaded to a searchable database [4], and from a mix of public and private cameras [1][3].
One next step
Search for your own city’s plate reader policy this week. It takes a few minutes, and it tells you who is watching the road you drive on every day. Then get back to your life, and keep a dashcam running if you can.