Essex Police has suspended the deployment of live facial recognition (LFR) cameras following a pivotal study revealing disproportionate identification rates of Black individuals compared to other ethnic groups. This development raises urgent questions about fairness and accuracy in cutting-edge law enforcement technology.
Facial Recognition Vans Under Scrutiny
The LFR cameras, mounted on mobile vans, scan crowds in real time to cross-check faces against police watchlists, aiming to identify wanted individuals swiftly. By the close of last year, thirteen police forces across the UK had adopted this technology. The Home Secretary announced plans in January to quintuple the number of LFR vans from 10 to 50, signaling a major expansion.
Despite this momentum, Essex Police paused its LFR operations after identifying a “potential bias in the positive identification rate.” The force has since collaborated with the algorithm’s software provider to update the system, believing these adjustments have addressed the issue.
University of Cambridge Study Reveals Bias
Researchers from the University of Cambridge enlisted nearly 200 volunteers to test the system during Essex Police’s deployment. Their findings showed the technology correctly recognized about half of those on the watchlist. False positives were rare, confirming the system’s overall reliability.
However, the study uncovered a troubling disparity: Black individuals were “statistically significantly more likely” to be correctly identified than people from other ethnic backgrounds. Men were also identified more frequently than women, spotlighting ongoing fairness concerns.
The researchers emphasized that these discrepancies “raise important questions about fairness that require continued monitoring.”
Essex Police Responds
Essex Police confirmed the Cambridge study was one of two commissioned. While the second study indicated no bias, the force decided to pause all LFR deployments to refine the algorithm further. They have since conducted additional academic assessments and now feel confident about resuming the technology’s use.
“We have revised our policies and procedures and are now confident that we can start deploying this important technology as part of policing operations to trace and arrest wanted criminals,” the police statement declared. “We will continue to monitor all results to ensure there is no risk of bias against any one section of the community.”
Evaluating Effectiveness and Fairness
The study also assessed the system’s overall impact since Essex Police began using LFR. Between August 2024 and February 2025, approximately 1.3 million faces were scanned, resulting in 48 arrests — roughly one arrest per 27,000 scans. Only one wrongful intervention was reported.
Researchers cautioned that different LFR systems and environmental factors could influence outcomes, stressing the need for ongoing testing to fully understand the technology’s performance across varied contexts.
Privacy, Oversight, and Future Challenges
As the UK government aims to expand LFR use, concerns about privacy and mass image collection remain at the forefront. The study underscored the necessity for “proportionality, transparency, and oversight” when deploying facial recognition.
The Information Commissioner’s Office (ICO) is actively monitoring the technology, urging all police forces to implement routine bias testing. The ICO warns that without rigorous scrutiny of technology design, training data, and watchlist composition, there is a “real risk of unfairness.”
The Home Office defends the technology, emphasizing that images are “immediately and automatically” deleted if they do not match watchlist entries. It stresses that all LFR deployments are targeted, intelligence-led, time-bound, and geographically limited.
Between January 2024 and September 2025, LFR contributed to over 1,300 arrests in London alone, involving serious offenses such as rape, domestic abuse, and grievous bodily harm (GBH).
