Flock Cameras Keep Getting It Wrong: False Hits, Gunpoint Stops, and a Woman Jailed for a Hit-and-Run She Didn’t Commit
- Rex Ballard
- 2 hours ago
- 5 min read
In Case You Needed Another Reason for Flock Cameras to Go
A Shasta Unfiltered reader from Tulsa, OK (raised in Redding and possibly returning) recently forwarded a ReclaimTheNet piece after reading our earlier coverage of Flock Safety cameras. The tip highlighted a Roseville, California, driver flagged six times because the system kept reading a “9” as an “8” on a never-stolen plate. That small OCR error is part of a much larger, documented pattern.
Flock claims high accuracy and more than 120,000 cameras nationwide. At the scale of tens of billions of monthly reads, even a small error rate produces a steady stream of false positives. Law enforcement agencies have treated those hits as actionable leads with real consequences for innocent people. Here are some of the clearest, reported cases of machine misreads, database failures, and failure to verify.
The Triple-Fatal Hit-and-Run That Nabbed the Wrong Woman
In October 2025, a speeding Dodge Durango caused a multi-vehicle crash on Interstate 4 in Volusia County, Florida, killing three people (including Flagler County Deputy Administrator Jorge Salinas and his wife) and seriously injuring another. Witnesses described a maroon Durango driven aggressively; one 911 caller gave partial plate characters that did not match the vehicle later targeted.
Florida Highway Patrol used a Flock camera capture of a black Durango near the scene to identify 23-year-old Lindsey Brooke Isaacs as the owner. She was arrested in April 2026 and charged with multiple counts of leaving the scene of a crash involving death and vehicular homicide. She spent 13 days in jail and faced the prospect of life-altering charges for months.
Her vehicle showed no crash damage consistent with the incident. Independent analysis by her attorney (a former traffic homicide investigator) showed timing and location evidence that placed her elsewhere. Prosecutors ultimately dropped all charges against Isaacs. The same day, they charged Alisa Lee Montalvo, whose maroon Durango matched the witness description and showed clear evidence of recent body work and concealment. Isaacs later sued, alleging false arrest and malicious prosecution.
A Flock hit near a crime scene is a lead, not proof of involvement, driver identity, or even the correct vehicle. In this case, basic verification of color, plate details, and damage was not done adequately before jailing an innocent woman.
Roseville’s 71% Problem
Closer to home, Roseville Police Department records from 2023–2024 show Flock sent 1,427 alerts for vehicles flagged as stolen or linked to felonies. Department analysis found the system misread the license plate in 71% of those alerts. Character confusions (9 for 8, 3 for 2, 1 for 4, and others) were common. One resident’s plate was misread at least six times; a dispatch supervisor noted they had essentially memorized the recurring false hit.
Roseville’s policy requires independent verification before a stop, so none of the false alerts produced contacts or arrests. Flock has attributed the high rate partly to the city’s atypical camera placement and older hardware (rear-only views mounted higher and farther than recommended). The company maintains overall accuracy is high under optimal conditions. Roseville’s experience still illustrates how often alerts can be wrong even in a controlled California deployment.
Other Documented Failures
The Institute for Justice and independent tallies have cataloged dozens of cases since 2018 in which Flock or similar ALPR alerts led to innocent drivers being stopped, detained at gunpoint, or jailed. Examples include:
San Diego (2025): Officers searching for a red Alfa Romeo linked to an attempted carjacking used Flock’s “vehicle signature” matching (make, model, color, and other traits). It flagged a different red Alfa Romeo five miles away. Occupants were arrested; one man spent nearly a month in jail.
Toledo, Ohio: A misread of “7” as “2” led to a stop at gunpoint and a police dog attack. The driver was injured, lost his job, and was later evicted; he settled a lawsuit.
Sherwood, Arkansas (2026): A digit misread flagged a couple’s SUV as stolen. Officers ordered them out at gunpoint while their six-week-old baby remained alone in the back seat.
Plymouth, Minnesota (2026): Automotive journalist Joel Feder, driving a loaner Range Rover with manufacturer plates, was boxed in by multiple squad cars. A California stolen-plate entry with incomplete digits, combined with the system’s handling of non-standard plate fonts, triggered tracking for days.
Colorado porch-theft case: Chrisanna Elser was accused after Flock data placed her Rivian in the area around the time of a package theft. An officer treated the combination of camera hits and doorbell video as conclusive. She had to gather her own dashcam, phone location, and other evidence to clear herself; the summons was voided.
Wisconsin (August 2026): Amber Newell was stopped at gunpoint twice in one week after a Flock alert linked her vehicle to a Milwaukee homicide. The alert should have been cleared days earlier; staff failed to remove it.
Common failure modes are OCR confusion of similar characters (especially when plates are dirty, angled, framed, or poorly lit), failure to clear resolved hotlist entries, incomplete data entry into national databases, and treating a “hit” as probable cause without visual confirmation of the actual plate or vehicle.
Accountability Gaps and Local Relevance
ReclaimTheNet’s reporting (the piece our reader shared) details additional problems with audit logs. Police departments have released poorly redacted spreadsheets that reveal who searched which plates. Flock itself sent guidance to customers in September advising redaction of plates, search reasons, and case numbers from the very logs marketed as accountability tools. Records requests often get pushed into a vendor process with tight windows and pressure to narrow scope. A hit in an audit log is only a starting point; confirmation requires a properly scoped public-records request that many people never successfully complete.
Shasta County has its own 24 Flock cameras (installed in early 2026, funded in part by Zogg Fire settlement money). The Sheriff’s Office has reported successes—stolen vehicle recoveries, felony arrests, and Silver Alerts. Local transparency portals and policy statements emphasize no facial recognition and California restrictions on certain uses. Those claims of utility do not erase the documented national error rate or the human cost when verification fails.
Technology that generates leads at massive scale is useful only when agencies treat alerts as starting points requiring independent confirmation, clear hotlists promptly, and maintain transparent, usable audit trails. California still lacks a statewide requirement that officers independently verify ALPR hits before initiating stops (unlike some other states). Without that discipline, the system continues to put innocent drivers—including a woman jailed for a fatal hit-and-run she did not commit—in the crosshairs of armed encounters and criminal charges.
Readers who want to check whether their own plate has been searched can start with the methods outlined by ReclaimTheNet and with local public records processes. The pattern of errors is no longer theoretical. It is documented, recurring, and already affecting real people in California and beyond.
