Britain's Biggest Police Force Scanned 1.7 Million Faces This Year

Multiple CCTV surveillance cameras mounted on a pole against a grey sky

TL;DR: The Metropolitan Police scanned more than 1.7 million faces with live facial recognition (LFR) in the first four months of 2026, an 87% increase over the same period last year ^[1][2]. The UK's biometrics commissioners for England, Wales, and Scotland have issued a joint warning: the legal framework governing the technology is a "patchwork" that is failing to keep pace with deployment ^[3]. Researchers at Queen Mary University of London found an 81% error rate across six Met LFR trials: only 8 of 42 matches were correct ^[2][4]. The technology has already led to wrongful arrests with documented racial bias: false match rates of 5.5% for Black faces and 4.0% for Asian faces, versus 0.04% for white faces ^[5]. The Home Office just committed £115 million to a new Police.AI centre to roll the technology out to all 43 forces in England and Wales, but legislation to govern it is at least three years away ^[1][2].

The Numbers Are Moving Faster Than the Law

In 2025, the Met Police conducted live facial recognition deployments in central London on a regular but limited basis. In 2026, the programme is accelerating hard.

The Met now runs multi-hour LFR sessions in high-traffic London locations, scanning every face that walks past a camera and comparing it against a watchlist of wanted individuals ^[1]. Behind the live deployments, officers run more than 25,000 retrospective facial recognition searches per month, matching images from CCTV, social media, and other footage against police databases ^[2].

Each of the 43 police forces in England and Wales currently sets its own AI procurement policy. There is no national standard for how facial recognition is deployed, what accuracy threshold triggers an alert, or how long the data is retained ^[3].

Some forces use a similarity threshold of 0.6. The National Physical Laboratory recommends 0.64, which the Home Office adopted from July 2024. Essex Police uses a similarity score of 55. Nobody is checking ^[3].

81% Wrong

The accuracy question is not theoretical. Researchers at Queen Mary University of London studied six Met live facial recognition trials and found that only 8 of 42 matches were correct: an 81% error rate ^[2][4].

The Met disputes this framing. They measure accuracy against total faces processed, meaning if a camera scans 10,000 people and makes 42 alerts, only 34 of which are wrong, the Met calls that a 0.34% error rate. The mathematically honest version: if you are flagged by this system, there is an 81% chance it got you wrong.

Dr. Brian Plastow, Scotland's biometrics commissioner, put it bluntly: facial recognition is "nowhere near as effective as the police claim it is" ^[3].

Prof. William Webster, commissioner for England and Wales, said the "slow pace of legislation was trying to catch up with the real world" ^[3]. The commissioners' joint position: police are "marking their own homework" on accuracy and deployment ^[2].

When the Algorithm Gets It Wrong, It Gets It Wrong by Race

Alvi Choudhury is a 26-year-old software engineer from Southampton. On January 7, 2026, officers from Hampshire Constabulary arrived at his parents' home and arrested him for a burglary at the Milton Keynes Buddhist Vihara, a crime committed 100 miles away by someone who looks nothing like him ^[5].

Thames Valley Police's retrospective facial recognition system matched Choudhury's mugshot (taken during a prior false arrest in 2021) to CCTV footage of the actual thief. Despite Choudhury having a beard and being visibly older than the suspect, officers took the algorithmic match at face value. He sat in custody for ten hours before detectives realised the mistake ^[5].

Home Office-commissioned research, disclosed in December 2025, showed why this keeps happening. At certain settings, the technology produces false positive rates of 5.5% for Black faces and 4.0% for Asian faces, compared to 0.04% for white faces. That is a 137-fold difference ^[5].

Thames Valley Police wrote to Choudhury afterward, acknowledging his arrest "may have been the result of bias within facial recognition technology", something the force knew to be a "wider issue" ^[5]. Choudhury is now suing.

He is not alone. Shaun Thompson, a 39-year-old anti-knife-crime campaigner, was wrongly flagged by the Met's LFR cameras at London Bridge while returning from a community patrol. Officers detained him for over 20 minutes, demanded fingerprints, and threatened him with arrest, despite him providing multiple forms of ID ^[6]. Thompson and Big Brother Watch director Silkie Carlo challenged the Met's LFR policy in the High Court.

The High Court Said It Was Fine

On April 21, 2026, Lord Justice Holgate and Mrs Justice Farbey dismissed the challenge, ruling the Met's facial recognition system lawful and compatible with human rights. The judges found the Met's policy contains "clear, interlocking and cumulative constraints" ^[6].

Thompson and Carlo are appealing.

"This is a disappointing judgment but the fight against live facial recognition mass surveillance is far from over," Carlo said ^[6].

The case did force one concrete change: the Met revised its LFR policy, and Thompson received a payment for the wrongful identification ^[6].

The Retail Pipeline

It is not just the police. Facewatch, a private facial recognition company used by retailers across the UK, sends 500,000+ alerts annually to shops ^[3]. Big Brother Watch has received reports from 21 people wrongly placed on watchlists and flagged every time they enter participating stores ^[3].

A retired health-and-safety professional named Ian Clayton was removed from a shop after Facewatch flagged him as a false positive ^[2]. Liberty has documented cases of LFR being used to track children as young as 12 ^[2].

The former Biometrics and Surveillance Camera Commissioner, Fraser Sampson, now sits on Facewatch's board as a non-executive director ^[3]. The watchdog turned vendor.

£115 Million to Scale Up

The Home Office is not slowing down. It has committed £115 million to a new Police.AI centre that will standardise facial recognition deployment across all 43 forces in England and Wales ^[1][2].

The goal is clear: what the Met is doing in London, every force should be doing everywhere. But the commissioners say the governance framework to support that does not exist yet.

The Crime and Policing Bill, currently moving through Parliament, is expected to include some governance provisions. The Home Office has confirmed it is considering a new legal framework, the first signal that primary legislation, rather than guidance, may follow ^[2]. But both commissioners say that legislation is at least three years away ^[1][2].

Three years of unchecked expansion, with an 81% error rate, documented racial bias, and a regulator that has postponed its own audit indefinitely. The ICO was scheduled to audit the Met's facial recognition programme in October 2025. That audit has been postponed with no new date ^[2].

57% of Britons See This Coming

Polling by Face Int found that 57% of Britons view live facial recognition as "another step towards turning the UK into a surveillance society." Another 62% worry about being misidentified ^[2]. Given the 81% error rate and the racial bias data, their worry is not paranoia. It is math.

Scotland may act first. Dr. Plastow has been pushing for Scotland to become the first nation to pass primary legislation specifically covering live facial recognition, rather than relying on the patchwork of existing data protection and human rights law ^[3]. Whether Holyrood moves before Westminster is an open question.

What You Can Do

Know your rights: In the UK, you are not legally required to walk past a live facial recognition camera. Deployments are usually signposted: if you see one, you can change your route. The Met says participation is voluntary, though in practice, busy streets make avoidance difficult.
Support the appeal: Big Brother Watch is crowdfunding the Thompson v. Met Police appeal. A successful appeal could set binding precedent on LFR's compatibility with human rights law.
Check the Facewatch map: If you have been denied entry to a shop or treated suspiciously, you may be on a Facewatch list. Submit a Subject Access Request under UK GDPR to both the retailer and Facewatch to find out what data they hold on you.
Contact your MP: The Crime and Policing Bill is the immediate legislative vehicle. Ask your MP whether it will include mandatory accuracy thresholds, independent oversight, and a prohibition on use against children.
File a complaint: If you believe you have been wrongly identified, contact the ICO. Every complaint adds to the regulatory record that the commissioners are building.

Sources

Published: May 12, 2026