Police Expansion of Live Facial Recognition Cameras: A Shift Towards ‘General’ Surveillance

Authorities anticipate that live facial recognition cameras may soon be “prevalent” across England and Wales, as indicated by internal documents revealing nearly 5 million face scans conducted last year.

A joint investigation by the Guardian and Liberty investigates showcases the rapid integration of this technology into UK law enforcement practices.

The government is simplifying police access to a wide range of image repositories, including passports and immigration databases, for past facial recognition searches, alongside significant financial investments in new hardware.

Live facial recognition entails real-time identification of faces captured by surveillance cameras, compared against a police watch list.

Conversely, retrospective facial recognition software allows police to match archived images from databases with those recorded on CCTV or similar systems.

The implementation of this technology is believed to be widespread in urban areas and transportation hubs across England and Wales, as noted in funding documents produced by South Wales Police and shared by the Metropolitan Police under the Freedom of Information Act.

The inaugural fixed live facial recognition camera is set to be trialed this summer in Croydon, located south of London.

This expansion comes despite the absence of any mention of facial recognition in the relevant congressional legislation.

Critics contend that police are permitted to “self-regulate” this technology, while there have been instances where previous algorithms disproportionately misidentified individuals from Black communities.

Following a 2020 Court of Appeals ruling that deemed South Wales Police’s live facial recognition practices unlawful, the Police College issued guidance emphasizing that “thresholds must be carefully set to enhance the likelihood of accurate alerts while keeping false alert rates within acceptable limits.”

There remains no statutory framework directing the standards or technology applied in this context.

Earlier this month, Police Minister Diane Johnson informed Congress that “we must evaluate whether a tailored legislative framework is necessary to govern the deployment of live facial recognition technology for law enforcement,” but further details from the Home Office are still pending.

Facial recognition cameras have been tested in London and South Wales since 2016; however, the pace at which police have adopted this technology has surged over the past year.

A survey conducted by the Guardian and Liberty revealed:

• Last year, police scanned nearly 4.7 million faces using live facial recognition cameras—over double the figures from 2023. Data indicates that a minimum of 256 live recognition vans were operational in 2024.

• Mobile units comprising 10 live facial recognition vans can be dispatched anywhere in the UK within a matter of days to bolster national capabilities, with eight police forces having deployed this technology, while the Met has four such vans.

• Authorities are exploring a fixed infrastructure to establish a “safety zone” by deploying a network of live facial recognition cameras throughout London’s West End. Met officials indicated that this remains a viable option.

• The force has nearly doubled the number of retrospective facial recognition searches on the National Police Database (PND) from 138,720 in 2023 to 252,798. The PND contains administrative mug shots, including many held unlawfully for individuals not formally charged or convicted of any offenses.

• Over the past two years, more than 1,000 facial recognition searches have utilized the UK passport database, with officers increasingly accessing 110 matches from the Home Office immigration database last year. Authorities concluded that using a passport database for facial recognition “presents no risk.”

• The Home Office is collaborating with the police to develop a new national facial recognition system termed strategic facial matchers, which will enable searches across various databases, including custody images and immigration records.

Lindsey Chiswick, Met’s Intelligence Director General and the National Police Chiefs’ Council’s Facial Recognition lead, stated that five out of five London residents support the police’s utilization of advanced technologies like facial recognition cameras, based on the survey findings.

Recently, registered sex offender David Chenelle, a 73-year-old from Lewisham, was sentenced to two years after being caught alone with a 6-year-old girl through live facial recognition technology. He had previously served nine years for 21 offenses involving children.

In 2024, the Met arrested 587 individuals, with 424 of those arrests backed by live facial recognition technology, leading to formal charges.

Among those arrested, 58 registered sex offenders faced serious violations of their conditions, with 38 subsequently charged.

Chiswick noted: “Given the limited resources and time available, the demand is high, and we see criminals exploiting technology on an expansive scale.

“There’s a chance for law enforcement to evolve. Discussions about leveraging AI are abundant, but we must embrace the opportunities presented by technology and data.”

Chiswick emphasized that the Met’s approach is to “proceed cautiously and evaluate at each phase,” while noting that “there may be advantages to some form of framework or statutory guidance.”

The MET employs facial recognition cameras in contexts aimed at ensuring statistical significance regarding gender or ethnic bias in misidentification instances.

Chiswick remarked: “I refuse to utilize biased algorithms in London. Each instance carries weight. The government raises concerns: Is there no issue regarding artificial intelligence?

“When selecting an algorithm’s purchaser, determining the training data employed, and assessing the origin of the technology, testing it thoroughly is paramount; you are obliged to operate within a specific context.”

The Ministry of Home Affairs did not provide a comment upon request.