Flock Lock-in Undermines City Oversight
LAPD's 5-Year Data Retention Clause
The Los Angeles Police Department (LAPD) allowed its contract to expire, but an audit revealed a clause permitting Flock to retain data for five years for "any purpose," including training machine learning algorithms, a period 60 times longer than Flock's standard policy, CNET and The Hill reported. Aurora, Colorado, paid a $1.9 million settlement after the system failed to verify the state of origin for a stolen vehicle alert, NPR, a San Clemente government document, and Emeraldbook.org documented. Stateline, CNET, and Fresno government documents show that the LAPD subsequently suspended camera use and drafted new language to prohibit AI data training. NBC News and Stateline observed that Flock's "Evidence Mode" allows specific license plate data tied to a case number to be preserved indefinitely. While Flock changed its default data retention policy from 30 days to a recommended 7 days, existing customers can maintain longer periods, ASIS International, Stateline, and NBC News explained. Surveillant.ai found that cloud storage is bundled into the annual lease, not priced separately based on retention duration, which diminishes the default's role as a mandatory data purge mechanism.
Flock's Black Box AI Algorithms
A University of Washington study, NPR, and Democracy Now explain that AI-driven interpretation automatically reads images and categorizes vehicles by various attributes, such as license plate, make, model, color, and unique features. A City of Frontenac document noted that this interpretation is managed entirely by the vendor, meaning municipalities must accept these proprietary algorithmic categorizations rather than independently verifying how vehicles are tagged, flagged, or searched. The algorithms operate as a black box, preventing municipalities from independently verifying algorithmic bias or the accuracy of search results without access to the underlying training data, Democracy Now and NBC News reported. A University of Washington study determined that Flock has reportedly refused to allow independent security analysts to test its devices to verify company claims about AI accuracy and data processing.
Roseville's 71% Alert Failure Rate
Roseville, California, documented a 71% failure rate in Flock alerts for stolen vehicles and felonies between 2023 and 2024, according to CNET, a preprint paper, a Fresno government document, and Surveillant.ai. This led to a policy requiring officers to independently verify license plates before acting on alerts, as CNET, a preprint paper, a Fresno government document, and Surveillant.ai confirmed. A University of Washington study reported that Morristown, Tennessee, settled a case after an "O" was misread as a "0." Toledo, Ohio, settled for $35,000 after a "7" was misread as a "2," and Atherton, California, paid $45,000 following an "H" misread as an "M," a University of Washington study detailed. The same University of Washington study also noted that Española, New Mexico, settled a case in 2025 involving a "2" misread as a "7." Richmond, Virginia, identified racial and class bias in camera placement, remediating this by banning federal agency access to Automatic License Plate Recognition (ALPR) data, Emeraldbook.org asserted.
Flock's Audit Tools Lack Search Results
A University of Washington study found that the ACLU has stated there is "no evidence that Flock's audit assistance tool works consistently to address police misuse" and calls for independent evaluators, which highlights the difficulty for local agencies to determine the actual scope and impact of data access. The study further observed that proprietary API restrictions also function as a structural bottleneck, preventing bulk data extraction necessary for meaningful full-dataset statistical auditing of retention compliance and search patterns. Flock has introduced configurable audit tools, including a mandatory Audit Assistance tool by the end of 2026 to flag abnormal search activity and require case codes for queries, ASIS International, Stateline, and NBC News reported. However, these tools primarily offload accountability verification onto understaffed local administrators, according to a University of Washington study.
Flock's Architecture Creates Data Dependency
Flock's proprietary architecture creates a structural data dependency that functionally undermines municipal accountability, despite contractual data ownership. This reliance limits local governments' ability to independently verify algorithmic accuracy, audit for bias, or control data sharing, potentially leading to unchecked surveillance expansion and privacy infringements. The documented instances of misreads and unauthorized data sharing show the risks of concentrating interpretive decision-making within a proprietary black-box system.
Comments ()