Dozens of Innocent Motorists Have Been Pulled Over, Detained at Gunpoint, or Jailed Due to AI License Plate Camera Errors

In February of this year, a Flock camera in Sherwood, Arkansas, misread the license plate of an SUV, leading officers to detain an innocent couple at gunpoint while their six-week-old baby sat alone in a car seat in the back of the vehicle.

“Those cameras are placed everywhere, and they hit license plates,” one of the officers explained as they uncuffed the couple after realizing their error several minutes later. “I’m not gonna say they’re completely perfect, because, you know, that’s modern technology.” 

As the number of automated license plate readers (ALPRs) has skyrocketed nationwide, errors like this one are becoming increasingly common. An Institute for Justice (IJ) review of media reports and court records finds there have been at least 24 such cases since 2018, with the majority of those happening since 2023. 

What happens when innocent people are stopped 

In recent months, stories have proliferated about innocent motorists enduring the indignity of being pulled over repeatedly until police departments figure out the root cause of the ALPR errors. But those drivers have it relatively easy: in nearly two-thirds of the cases IJ analyzed, officers did not realize their error until after they had drawn and pointed their guns at innocent people. 

“Every one of those stops is a high-risk encounter where a wrong move, a misunderstanding, or a moment of fear can turn deadly,” said Michael Soyfer, an IJ attorney who is representing residents of San Jose and Norfolk in lawsuits challenging their cities’ ALPR surveillance networks. “No one should have to prove their innocence on the side of the road because a camera couldn’t tell a zero from an O.” 

The recent uptick in ALPR errors owes largely to Flock Safety, which in the past several years has become the leading ALPR provider in the market. The company claims its cameras accurately capture 93 out of every 100 license plates that pass by them.  

It all adds up to more than 20 billion individual license plate readings each month, according to Flock’s marketing materials. Even taking Flock’s claims at face value, a 93% accuracy rate implies well over one billion inaccurate readings per month, which is especially concerning given that the vast majority of plates scanned are not relevant to an investigation. 

Those machine errors happen for a variety of reasons. Minor obstructions like dirt or license plate holders can sometimes throw off the devices’ character recognition, for instance. And the machines often have trouble distinguishing between visually similar characters, like 0 and O, or 2 and 7. 

How human errors contribute 

But those machine errors account for about one-third of the mistakes in the cases IJ identified. The rest involved human error, with officers entering wrong information into the system or misinterpreting what the ALPR data says.  

Last year in San Diego, for instance, officers were searching for a red Alfa Romeo connected to an attempted carjacking. The officers didn’t have a plate and were instead relying on Flock’s “vehicle signature” technology, which captures detailed characteristics of individual cars like make, model, and color. 

The Flock system gave them a positive hit on a superficially matching car—but it was a totally different red Alfa Romeo, located five miles away from the crime at the time it occurred. Officers nevertheless arrested all three of the car’s occupants. One passenger spent nearly a month behind bars during the holidays before officers realized their error and set him free. 

In other cases, officers have entered multiple variations of the same license plate into the system, in part to account for ALPR character recognition errors. 

“The Constitution requires real suspicion before the government can seize someone at gunpoint, and a computer hit that no one bothered to confirm doesn’t come close,” Soyfer said. 

The list below includes only cases involving stationary ALPRs like Flock’s. Cases of mistaken identity involving vehicle-based ALPRs, traffic cameras, or toll cameras are not included. The review builds on previous efforts to tally ALPR errors, including those by CBS News and the Electronic Frontier Foundation. 

These 24 cases are most likely an undercount, perhaps by a significant margin. There is no comprehensive database of ALPR errors, and IJ has identified just one jurisdiction that publishes regular reports of such errors: Oak Park, Illinois. Those show that, in a typical reporting period, one-third or more of the traffic stops prompted by a Flock alert resulted in the drivers being released due to data problems.  

“There is no evidence whatsoever that Flock Safety ALPRs have played a meaningful role in any Oak Park crime investigation since their installation in 2022,” the oversight board publishing those reports concluded in 2025. 

Oak Park ended its contract with Flock in August of that year.

List of known cases