Blab Tech
Police Stop Highlights Flaws in Flock Safetys License-Plate Surveillance System
A journalist in Plymouth, Minnesota, found himself surrounded by four police cars and taken to a station after a Flock Safety automated license‑plate reader (ALPR) mistakenly flagged his vehicle as a stolen‑plate suspect. The incident, captured on body‑camera footage that has since gone viral, highlights growing doubts about the accuracy and oversight of AI‑powered surveillance used by law‑enforcement agencies.
Joel Feder, editor‑in‑chief of the automotive news site The Drive, was driving a Range Rover with the registration 34 10 DTM when police in the Kohl’s parking lot believed the plate was stolen. According to a podcast interview, the officers had received an alert from Flock’s system two days earlier that the same plate had been reported stolen in Los Angeles. When Feder approached the police, the officers boxed and pinned him, a procedure they justified by citing the plate’s status as a “stolen” record.
The error stemmed from a data entry mistake in the National Crime Information Center (NCIC) database. The Los Angeles sheriff’s department had logged a stolen plate as 34 DTM, omitting the middle digit “10.” Flock’s ALPRs pull data from NCIC and issue alerts when a plate matches a hot‑list entry. In this case, the system matched the incomplete record and sent an alert to the Plymouth officers, who then acted on the incomplete data.
Founded in 2017, Flock Safety operates more than 5,000 cameras across 49 U.S. states and processes roughly 20 billion license‑plate scans per month. The company markets its network to police departments, neighborhood associations, and private property owners. In addition to ALPRs, Flock offers video, gun‑shot detection, and, through a partnership with defense contractor Leonardo, the SignalTrace technology that scans Bluetooth and Wi‑Fi signals to build digital maps of individuals’ movements.
When contacted, Flock’s chief communications officer Josh Thomas apologized for the incident and explained the system’s logic. “The system was asked to look for 34 DTM,” Thomas said. “It found that plate, sent the alert, and the police had to verify the match.” Thomas acknowledged that the system does not flag partial matches as errors and that human verification is required. He said Flock is working with the FBI’s NCIC division to add contextual information to alerts so officers can see when a plate is incomplete.
Plymouth Police Chief Eric Fadden confirmed that the officers followed protocol and did not draw their weapons. He noted that the error was due to the NCIC data entry mistake and that the officers had no reason to suspect an armed suspect once the vehicle was identified as a Range Rover. Fadden said the department would review its procedures for handling alerts from private vendors.
The incident has fed into a broader debate about the proliferation of ALPRs and other surveillance technologies. Several states have cut or paused contracts with Flock after public outcry, and the company has recently ended a pilot for acoustic gun‑shot detection devices. Critics argue that the technology’s high error rate and lack of oversight create a “dragnet” that can trap innocent citizens.
At the time of the interview, the Plymouth police report had not yet been corrected, and Flock’s contract with the Los Angeles sheriff’s department had not been renewed. City council meetings in Plymouth and other jurisdictions are ongoing, with officials calling for clearer regulations on data retention, accuracy, and vendor accountability.
The incident illustrates the need for better data quality controls and regulatory oversight of private surveillance vendors. While Flock has pledged to improve its alert system, the case shows that human verification remains essential and that errors in public databases can have immediate, real‑world consequences.
As law‑enforcement agencies continue to deploy AI‑driven tools, the balance between public safety and civil liberties will remain a contentious issue. The outcome of the Plymouth case may influence how other cities negotiate contracts with vendors like Flock and how state and federal laws evolve to address the challenges of mass surveillance.
Joel Feder, editor‑in‑chief of the automotive news site The Drive, was driving a Range Rover with the registration 34 10 DTM when police in the Kohl’s parking lot believed the plate was stolen. According to a podcast interview, the officers had received an alert from Flock’s system two days earlier that the same plate had been reported stolen in Los Angeles. When Feder approached the police, the officers boxed and pinned him, a procedure they justified by citing the plate’s status as a “stolen” record.
The error stemmed from a data entry mistake in the National Crime Information Center (NCIC) database. The Los Angeles sheriff’s department had logged a stolen plate as 34 DTM, omitting the middle digit “10.” Flock’s ALPRs pull data from NCIC and issue alerts when a plate matches a hot‑list entry. In this case, the system matched the incomplete record and sent an alert to the Plymouth officers, who then acted on the incomplete data.
Founded in 2017, Flock Safety operates more than 5,000 cameras across 49 U.S. states and processes roughly 20 billion license‑plate scans per month. The company markets its network to police departments, neighborhood associations, and private property owners. In addition to ALPRs, Flock offers video, gun‑shot detection, and, through a partnership with defense contractor Leonardo, the SignalTrace technology that scans Bluetooth and Wi‑Fi signals to build digital maps of individuals’ movements.
When contacted, Flock’s chief communications officer Josh Thomas apologized for the incident and explained the system’s logic. “The system was asked to look for 34 DTM,” Thomas said. “It found that plate, sent the alert, and the police had to verify the match.” Thomas acknowledged that the system does not flag partial matches as errors and that human verification is required. He said Flock is working with the FBI’s NCIC division to add contextual information to alerts so officers can see when a plate is incomplete.
Plymouth Police Chief Eric Fadden confirmed that the officers followed protocol and did not draw their weapons. He noted that the error was due to the NCIC data entry mistake and that the officers had no reason to suspect an armed suspect once the vehicle was identified as a Range Rover. Fadden said the department would review its procedures for handling alerts from private vendors.
The incident has fed into a broader debate about the proliferation of ALPRs and other surveillance technologies. Several states have cut or paused contracts with Flock after public outcry, and the company has recently ended a pilot for acoustic gun‑shot detection devices. Critics argue that the technology’s high error rate and lack of oversight create a “dragnet” that can trap innocent citizens.
At the time of the interview, the Plymouth police report had not yet been corrected, and Flock’s contract with the Los Angeles sheriff’s department had not been renewed. City council meetings in Plymouth and other jurisdictions are ongoing, with officials calling for clearer regulations on data retention, accuracy, and vendor accountability.
The incident illustrates the need for better data quality controls and regulatory oversight of private surveillance vendors. While Flock has pledged to improve its alert system, the case shows that human verification remains essential and that errors in public databases can have immediate, real‑world consequences.
As law‑enforcement agencies continue to deploy AI‑driven tools, the balance between public safety and civil liberties will remain a contentious issue. The outcome of the Plymouth case may influence how other cities negotiate contracts with vendors like Flock and how state and federal laws evolve to address the challenges of mass surveillance.
