The flashing blue lights came without warning.

According to the Institute for Justice, a public interest law firm, an Arkansas family was driving with their 6-week-old baby when police officers surrounded their SUV, weapons drawn, after an automated license plate reader allegedly misread a single character on their license plate. The vehicle wasn’t stolen. The family wasn’t wanted. But by the time officers realized the mistake, the encounter had become another example cited by critics who warn that AI-powered policing tools can have serious consequences when they fail.

The Institute for Justice identified at least two dozen documented cases since 2018 in which innocent motorists were allegedly pulled over, detained at gunpoint or arrested after automated license plate reader systems generated erroneous alerts or officers acted on inaccurate information associated with those alerts. The organization said roughly one-third of those cases involved technology errors, such as cameras confusing similar characters on license plates, while the remaining incidents stemmed from human mistakes, including incorrect database entries and failures to remove recovered vehicles from law enforcement hot lists.

“Every one of those stops is a high-risk encounter where a wrong move, a misunderstanding, or a moment of fear can turn deadly,” Institute for Justice attorney Michael Soyfer said in the report. “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 cases have become a rallying point for privacy advocates, who argue the debate over automated license plate readers is no longer limited to whether the cameras help solve crimes.

That argument is also playing out in a lawsuit filed by the New York Civil Liberties Union on behalf of four Westchester County, New York, residents. The lawsuit challenges one of the nation’s largest automated license plate reader networks, alleging the system has evolved from an investigative tool into a form of mass surveillance capable of documenting the movements of millions of people who are not suspected of criminal activity.

According to the complaint, the Westchester County Police Department operates at least 575 automated license plate readers that recorded more than 264 million vehicle sightings in 2024 alone. More than 99% of those records, the lawsuit alleges, were unrelated to any criminal investigation when they were collected. Yet the information was retained in a searchable database for at least two years.

“The information collected in this dragnet amounts to a digital dossier on the comings and goings of named plaintiffs and all other drivers in Westchester County,” the complaint states. It argues the technology allows police to reconstruct people’s daily routines, political activities, religious practices, medical visits and personal associations.

The plaintiffs do not allege they were wrongfully stopped because of the technology. Instead, they contend their ordinary lives have been quietly cataloged without their knowledge or consent.

One plaintiff, retired public health professional Yukiko Ann Umemoto, learned through a public records request that her vehicle had been recorded at least 1,134 times. According to the lawsuit, those records documented routine trips to grocery stores, medical appointments, volunteer activities, photography classes, political canvassing events and public demonstrations, creating what the complaint describes as “a mosaic” of her life.

The lawsuit also questions the technology’s accuracy. It cites research finding automated license plate reader systems misidentified passing vehicles in more than one-third of sampled cases and alleges similar inaccuracies appeared in Westchester’s own records. According to the complaint, the system incorrectly identified plaintiff Leah Nelson’s vehicle color in 95% of recorded reads and repeatedly misclassified Umemoto’s sedan as a truck, motorcycle and tractor-trailer. The complaint argues those mistakes “amplify the risk that Westchester drivers will be wrongfully stopped or investigated.”

Beyond simply reading license plates, the lawsuit alleges modern systems increasingly rely on AI to analyze years of vehicle data, identify travel patterns and determine which vehicles frequently travel together.

Law enforcement agencies, meanwhile, argue automated license plate readers have become indispensable.

Departments across the country credit the systems with helping investigators recover stolen vehicles, identify suspects, locate missing people and generate leads that might otherwise be impossible to develop. Supporters contend the cameras do not identify people, only vehicles, and say they allow investigators to work more efficiently while deterring crime.

Manufacturers also dispute the suggestion that the technology alone should determine police action. Flock Safety, one of the nation’s largest license plate reader providers, has repeatedly said its alerts are intended to serve as investigative leads rather than probable cause for a stop or arrest. The company says officers are expected to independently verify alerts before taking enforcement action and that proper policies, auditing and training are essential to responsible use.

As automated license plate reader networks continue to expand nationwide, the technology sits at the intersection of two competing realities. To investigators, the cameras can quickly connect vehicles to crimes and accelerate investigations that once took weeks. To critics, the same systems create an unprecedented record of where innocent people live, work, worship, seek medical care and spend their daily lives.