Roseville, a Sacramento suburb, has used Flock Safety's AI cameras since 2021. The records obtained by Business Insider through public-records requests, including emails, data, contracts, and meeting notes, show a long series of Flock errors. In summer 2024, an officer called a resident about a recurring plate problem: Flock's system had at least six times mistaken his plate and categorized his car as stolen or tied to a felony.
The department saw the software was reading his "9" as an "8"; the owner said he would remove his plate cover. Soon after, a similar alert appeared. A dispatch supervisor wrote to a colleague that it was "easier at this point we have it memorized."
Records also said cameras missed vehicles, sent blurry frames, confused plate characters and states, and delivered delayed alerts.
Officers have to check that a plate is stolen or have an independent reason to suspect a crime before stopping a car. NYU's Policing Project lists Montana, Virginia, Washington, and Kentucky as states that now require independent verification before a stop. California does not.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
California state Assembly Member Josh Lowenthal, who has pushed legislation to limit license-plate reader data retention, called Roseville's errors disturbing and said lawmakers should act. He said he would request a statewide review of reader reliability. "If the data is flawed, then so too will be our justice system," "Lowenthal told Business Insider. License-plate readers, he said", "should make Californians safer, not put them at risk."
Max Isaacs, the technology law and policy director at NYU's Policing Project, called the 71% misread rate "stunning." "In the absence of concrete rules that directly address the problem of misreads, this technology is not safe to be deployed in communities."
Flock says its plate-character reading accuracy exceeds 96% under optimal conditions. A Flock spokeswoman said Roseville requested a "particularly unique deployment": cameras capture only the backs of vehicles to avoid collecting faces and other identifiable personal details. Flock also said Roseville used older hardware, with cameras positioned higher and farther from traffic than the company usually recommends. Flock said it is "dedicated to going above and beyond" to help customers "even when the deployment is more challenging than our typical conditions." The company did not answer whether it runs its own accuracy checks on cameras.
Business Insider has already reported at least a dozen other cases in which plate misreads or officers' failure to verify a plate left innocent people facing guns or jail. In Toledo, Ohio, a police dog mauled Brandon Upchurch after a Flock camera misread his "7" as a "2." He said the injuries cost him his job and home; he settled with the city and an officer for $35,000. A 2021 IPVM test found Flock misidentified the state on about 1 in 10 plates; Flock later blocked IPVM from buying its cameras for testing.
More towns are freezing, pausing, or canceling their Flock agreements over privacy concerns, potential abuse, and information shared with federal immigration authorities. Roseville still uses Flock, saying the cameras have solved crimes, including cases that might never have been cleared. The city has spent about $450,000 on Flock products since 2020; its police budget last fiscal year was $69 million.
Roseville has reported issues to Flock dozens of times over four years, including recently, and still meets with the company twice a month. Flock maintains that Roseville's camera performance has greatly improved; the department disputes that. Roseville said its accuracy data for 2025 and 2026 is incomplete and it has not analyzed recent data "to verify to what degree performance has or has not improved."
"Our agency acknowledges that technology is not perfect and requires human oversight," the Roseville spokesperson said. "We take that responsibility seriously."
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets
