NewsBite analysis · AI, Security
CNET
Flock’s AI Cameras May Have Misread Over 70% of License Plates, Creating New Headaches for Cities
A study by the Roseville, California police department found that Flock Safety's AI surveillance cameras misread license plates more than 70% of the time. Critics argue this could lead to wrongful identification of vehicles, increasing the burden on law enforcement. Flock's claims of high accuracy do not reflect the issues encountered during real-world usage, highlighting significant reliability concerns that could impact their business with cities.
MonitorConfidence: HighShort term horizonReputational risk: High
Published 4 Aug 2026, 14:15 UTC3 min read
Visit original article ↗Lead image — CNET
Summary
TL;DR
- 71% failure rate in license plate recognition reported by Roseville police.
- Misreadings caused wrongful flagging of innocent vehicles as stolen.
- Flock claims 96% vehicle count accuracy, but lacks data on plate recognition.
- Cities are reconsidering or canceling contracts due to reliability issues.
- Placement and environmental factors heavily influence camera performance.
Interesting facts
- The study analyzed nearly 1,500 reports from Flock's cameras.
- Misidentification can happen when numbers are similar, e.g., mistaking a 9 for an 8.
- Flock's setup in Roseville was claimed to be atypical, though no details were provided.