Flock Safety Cameras Misread 71% of Alerts in California Town

In Roseville, none of the incorrect alerts led to a traffic stop or arrest, a department spokesman said. The city requires officers to verify a plate is stolen or have independent reasonable suspicion before making a stop. States including Montana, Virginia, Washington, and Kentucky have passed laws requiring such verification. California has not.

California Assembly Member Josh Lowenthal, who previously introduced legislation on license-plate reader data retention, called the Flock errors disturbing and said they require immediate legislative action. He said he would call for a statewide review of the reliability of the readers. "If the data is flawed, then so too will be our justice system," he said. The readers "should make Californians safer, not put them at risk."

Flock said Roseville's camera performance has improved significantly, a claim the police department disputed. Roseville said its data on alert accuracy is not complete for 2025 or 2026 and that it has not analyzed recent data to verify improvement. The company declined to say whether it independently tests its cameras' accuracy.

Roseville, which began using Flock cameras in 2021, continues to contract with the company because the cameras have helped solve crimes. The city has spent about $450,000 on Flock products since 2020. "Our agency acknowledges that technology is not perfect and requires human oversight," the Roseville spokesman said. "We take that responsibility seriously."

The cameras also missed vehicles entirely. Last fall, police trying to track a suspect in mailbox break-ins found that a Flock camera had missed the suspect's dark gray SUV. The same thing happened in January, during a search for a deli break-in suspect, and again the following day when a gray sedan was speeding and swerving.

Amanda Jones, a supervisor for the department's real-time crime center, documented missed and misread plates. In April 2025, she notified Flock that 16 of the 17 plates flagged the previous week were incorrect, including "blatant misreads" where a "3" was read as a "2." A lieutenant replied, "Ouch!"

Flock apologized and proposed solutions: reframing cameras, software updates to fix blurry plates and overexposed images. At one point, Flock's product director for machine learning suggested officers ask drivers to remove license plate frames that made it "very difficult for the machine vision to tell it is actually a 'E' and not an 'F.'" Roseville said it would not do that.

In 2024, Flock shared an analysis with Roseville outlining reasons for misreads: vehicles far from the camera were blurry, plates were cut off by trees or frames, and the software confused similar characters like "N" for "V" or "1" for "4."

Roseville is not the first to flag inaccuracies. In 2021, the research firm IPVM tested Flock's readers and found that Flock misidentified the state in about one of 10 reads and regularly misclassified vehicle type and make. IPVM said Flock subsequently blocked it from purchasing cameras for testing.

U.S. Sen. Alex Padilla of California told Business Insider that "inaccuracies stemming from law enforcement's use of this technology are deeply alarming, as are the mass surveillance capabilities that it enables." Technologies like Flock, he said, "should only be deployed with strong safeguards that protect both public safety and Americans' constitutional rights."