Essex police pause facial recognition camera use after study finds racial bias

2026-03-20 · Source: AI (artificial intelligence) | The Guardian · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Public Safety & Security, Public Policy & Governance · Depth: Intermediate, short

Summary

Essex police have suspended their use of live facial recognition (LFR) technology following a study by University of Cambridge academics that identified racial bias. The study, which involved 188 actors, found LFR cameras were "statistically significantly more likely to correctly identify black participants than participants from other ethnic groups," despite overall high accuracy for watchlist identifications and rare incorrect identifications. The Information Commissioner’s Office (ICO) revealed the pause, warning other forces to implement mitigations. This issue differs from concerns about misidentifying innocent people, such as a recent case where a man was arrested for a burglary 100 miles away due to retrospective face scanning software error. The government plans to increase LFR van availability five-fold, with 50 vans for every police force in England and Wales.

Key takeaway

For police forces and technology leaders deploying AI-enabled surveillance, this incident highlights the critical need for rigorous, independent bias testing before and during operational use. Your teams should proactively commission academic studies or third-party audits to identify and mitigate demographic biases in LFR systems. Ensure policies and procedures are revised based on these findings, and continuously monitor results to prevent disproportionate targeting of any community section, safeguarding public trust and legal compliance.

Key insights

Racial bias in live facial recognition systems can lead to disproportionate identification of specific ethnic groups.

Principles

• Bias in AI systems requires continuous monitoring.
• Algorithm overtraining can cause demographic bias.

Method

Academics deployed LFR cameras from marked police vans and observed 188 actors walking past to assess identification accuracy and demographic bias, publishing results to inform policy.

In practice

• Adjust system settings to rectify algorithm overtraining.
• Review LFR policies and procedures for bias risks.

Topics

• Live Facial Recognition
• Algorithmic Bias
• Racial Bias
• Police Technology
• AI Ethics

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Ethicist, Policy Maker, AI Operations Specialist

Related on AIssential

• How does live facial recognition work and how many UK police forces use it? — Live facial recognition is expanding in UK policing, raising efficiency versus civil liberties and bias concerns.
• Wanted an image of Educated and Uneducated Person, Made the mistake of asking copilot to make it. — AI models reflect societal biases present in their training data, often without understanding context.
• Edmonton Police Trial AI Facial Recognition Bodycams — The Edmonton Police Service's AI facial recognition bodycam trial is deemed Canada's highest-risk algorithmic surveillance program.
• Fairness Beyond Demographics: Optimizing Performance Across Appearance-Based Hidden Cohorts in Medical Imaging — Optimizing fairness across appearance-based hidden cohorts improves medical image analysis models without relying on demographic labels.
• Joy Buolamwini wins national contest for her work fighting bias in machine learning — Algorithmic bias, particularly in facial recognition, leads to discriminatory outcomes and requires active intervention.
• AI Bias in Hiring Tools: Real Disasters and How Teams Fixed Them — AI bias, often from skewed training data, leads to unfair outcomes, especially in critical applications like hiring.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI (artificial intelligence) | The Guardian.