Reference
ALPR error incident record
Documented cases where a plate reader alert led to an innocent person being stopped, detained, or jailed. Grouped by what actually went wrong, because that is what determines whether your agency is exposed to the same thing.
Reviewed 26 July 2026
About this page
The underlying case compilation is the work of the Institute for Justice, which reviewed media reports and court records and identified at least 27 such cases since 2018, most of them after 2023. Their full list, with a source link for every case, is published here and is the authoritative version. Go read it.
What I have added is the grouping. IJ organized their list chronologically. I have organized a selection of it by failure mode, because an agency reading this needs to know which of these could happen to them and which control prevents it. Those are different questions.
Two findings are worth stating plainly before the list. In close to two thirds of the cases IJ analyzed, officers did not realize the error until after they had drawn and pointed their weapons. And only about a third of the cases were camera errors. The rest came from people.
Character misread by the camera
The optical recognition returned the wrong character. These are machine errors, and they are the minority of documented cases, but they produce some of the worst outcomes because the alert looks completely normal to the officer receiving it.
| When | Where | System | What happened |
|---|---|---|---|
| Jun 2024 | Morristown, TN | Flock | A letter O was read as a zero. Two grandparents were held at gunpoint while their three year old granddaughter watched from the car. |
| Apr 2024 | Toledo, OH | Flock | A seven was read as a two. The driver was detained at gunpoint, a police dog was deployed on him, and he spent several hours in jail. |
| Feb 2026 | Sherwood, AR | Flock | A single character was misread. A couple was ordered out at gunpoint while their six week old infant stayed in the back seat. |
| Feb 2024 | York County, SC | Flock | A character was misread and the vehicle was flagged as stolen. The sheriff's office attributed the error to a tinted plate cover. |
| Jul 2023 | Espanola, NM | Flock | A digit was misread. Two sisters were detained at gunpoint. |
| Oct 2023 | Sedgwick County, KS | Flock | The system did not distinguish a temporary plate from a permanent one. Deputies detained an innocent couple at gunpoint. |
| Apr 2021 | Atherton, CA | unspecified | Dirt partially obscured the plate. The reader misread it and the driver was detained at gunpoint. |
Hotlist entry nobody cleared
The vehicle was legitimately flagged at some point. Then the case closed, or the car was recovered, and no one removed the record. The camera kept doing its job against a list that was no longer true. Every case in this group was preventable with a records procedure.
| When | Where | System | What happened |
|---|---|---|---|
| Nov 2018 | Hercules, CA | Vigilant | A stolen rental had been recovered but was never taken off the hotlist. The driver and his brother were detained at gunpoint. |
| Feb 2022 | Greenville, SC | unspecified | A recovered rental vehicle was still on a hot list. Two women were detained. |
| Mar 2023 | Jackson Township, OH | Flock | Officers failed to remove a plate from a hotlist. An innocent driver was pulled over and detained. |
Wrong plate entered by hand
Somebody typed a plate incorrectly, or entered several variations of a plate to compensate for expected reader errors, and an innocent person's plate ended up on an active list. The signature of this failure is repetition. The same driver gets stopped again and again until someone traces it back to the list.
| When | Where | System | What happened |
|---|---|---|---|
| Apr 2026 | Cherry Hills, CO | Flock | Officers mistakenly entered a driver's own plate on a hotlist. He was pulled over repeatedly. |
| Dec 2025 | Boulder County, CO | Flock | Same pattern. A driver's plate was entered on a hotlist by mistake and she was stopped repeatedly. |
| Aug 2025 | Jefferson County, CO | Flock | A third Colorado driver was repeatedly pulled over after her plate was mistakenly added to a hotlist. |
| Jul 2022 | Charlotte, NC | NDI | Incorrect information was entered into the system. An innocent woman was arrested. |
Data read wrong by the officer
The record was accurate. The interpretation was not. This group includes vehicle attribute matches treated as identifications, and alerts read as confirmations rather than leads. It is the group most directly addressed by verification training.
| When | Where | System | What happened |
|---|---|---|---|
| Nov 2025 | San Diego, CA | Flock | Officers relied on a vehicle signature match rather than a plate. The system matched a different car of the same make and color, five miles from the crime. One passenger spent nearly a month in jail. |
| Apr 2026 | Volusia County, FL | Flock | A capture was used to link the wrong car to a fatal accident. The driver was jailed for thirteen days before charges were dropped. |
| Aug 2025 | Redmond, WA | Flock | System data incorrectly connected a suspect's vehicle to his father. Officers detained the wrong man. |
| Sep 2025 | Denver, CO | Flock | An officer misread captures and attributed a series of thefts to an innocent woman. |
| Sep 2023 | Detroit, MI | unspecified | Officers misinterpreted reader data, detained an innocent woman at gunpoint, placed her autistic child in a patrol car, and impounded her vehicle for weeks. The city later settled. |
| Aug 2023 | Houston, TX | Flock | An alert for a stolen car was misinterpreted. Two innocent men were detained at gunpoint. |
| May 2024 | North Myrtle Beach, SC | unspecified | A vehicle was misidentified from an alert and an innocent teenager was detained at gunpoint. |
What this adds up to
Three of the four groups above have nothing to do with camera accuracy. Better hardware would not have prevented them. What would have prevented them is a records procedure, a review interval, and a habit of confirming a read before acting on it.
That is the whole argument for treating this as a governance problem rather than a technology problem. It is also why an agency can improve its position substantially without changing vendors or spending money.
A note on counting
IJ is explicit that 27 is almost certainly an undercount, and I agree. There is no national database of plate reader errors. Oak Park, Illinois appears to be the only jurisdiction that publishes regular reports of them. Everything else surfaces only through news coverage or litigation, which means the cases we can see are the ones where somebody sued or somebody talked.
If your agency tracked its own alert outcomes, you would know your own number. Almost none do. That is one of the things a program audit establishes.