A study of Flock Safety cameras in Roseville, California, found that most felony-related alerts were triggered by incorrect license plate readings . This data, spanning 2023 to 2024, suggests significant reliability issues within the automated surveillance system.
The 71% error rate in Roseville's felony alerts
According to an analysis of Roseville Police Department data, the system failed significantly when the stakes were highest. Out of 1,427 alerts examined, 71 percent of those involving stolen vehicles or felony-related reports were based on misread plates. These false positives represent a major breakdown in the technology's ability to accurately flag criminal activity.
The Roseville Police Department has reportedly invested approximately $450,000 in the Flock Safety system. While the department stated it has not yet made an arrest based on a faulty scan, the frequency of these errors means that law enforcement received incorrect notifications seven out of ten times in high-stakes categories, leading to a substantial waste of municipal resources.
Hardware height and privacy-focused scanning in California
Flock Safety has defended its technology by citing specific local deployment issues in Roseville. As reported by Gizmodo, a spokesperson for the company claimed that the city utilized older hardware and non-standard installation locations,placing cameras higher and further from vehicles than the company recommends.
The accuracy of the system may also have been impacted by local privacy preferences. To avoid capturing faces, Roseville officials requested that the Flock cameras only scan the rear of passing vehicles. This specific configuration, combined with the aforementioned hardware placement, may have contributed to the high rate of misidentification reported in the study.
A national pattern of 27 documented wrongful detentions
The implications of these errors extend far beyond the Sacramento suburb. While Flock Safety maintains a massive footprint—operating over 120,000 cameras across 49 states for more than 12,000 customers—other jurisdictions have already felt the impact of flawed data. At least 27 cases have been documented where drivers were mistakenly detained, pulled over, or arrested due to Flock camera alerts .
These incidents highlight a growing tension between the convenience of automated surveillance and the protection of civil rights. A notable example occurred in Arkansas, where a flawed alert led to a traumatic encounter for an innocent individual, illustrating the real-world danger of relying on unverified automated notifications.
The discrepancy between 96% OCR accuracy and high-stakes alerts
The Roseville findings raise critical questions regarding the gap between Flock Safety's internal testing and real-world performance . While the company claims an OCR accuracy of over 96 percent and a 99 percent capture rate in clear weather,these figures reflect all scans rather than just those that trigger criminal alerts. This leads to several unanswered questions : Does the system's accuracy plummet specifically when filtering for felony-related plates? To what extent did the 'rear-only' scanning mandate in Roseville degrade the software's performance? Finally, can the company prove that recent hardware upgrades have truly resolved the 71 percent error rate seen in previous years?
Comments 0