6°C

overcast clouds

TFL Updates
London Daily News

South Wales facial recognition figures raise questions over ethnic disparity

South Wales facial recognition figures raise questions over ethnic disparity

Ethnic minority people accounted for 24% of those selected for a particular type of police facial recognition check in South Wales in 2025, according to figures released following a freedom of information request.

The figure compares with a reported 7% ethnic minority share of the population in the force area. South Wales Police says the figures raise concerns about disproportionality and that it will continue to review its approach through its existing scrutiny arrangements.

The figures concern an officer-initiated mobile phone system. They do not describe everyone passing a camera, and they do not establish why the disparity occurred. A comparison with the whole population is a starting point for scrutiny, not a finding that any individual officer acted unlawfully.

How the mobile checks work

In an officer-initiated check, a police officer photographs an unknown person using a work-issued phone. The image is compared with a reference database to help the officer establish identity. The force says the database contains custody images held by South Wales and Gwent police and images of some missing people.

The force says officers may use the app to identify an unknown person they suspect is missing, at imminent risk of serious harm or wanted, when the person cannot or will not provide details, or is suspected of giving false details. It also describes uses involving people who are unconscious or have died, where establishing identity may help officers contact relatives.

Its guidance says officers should first try other reasonable, less intrusive ways to establish identity. A possible match requires an officer’s assessment; it is not, by itself, proof that someone has committed an offence. The force says images and biometric information captured through the app are automatically deleted from the app, while information about the circumstances and outcome is retained for scrutiny. It says body-worn video of use may also be retained for scrutiny and complaints.

This process is different from live facial recognition, which compares a camera feed with a watchlist, and from retrospective searches of an existing image during an investigation. The 2025 figures relate to the mobile checks rather than those other uses.

What the figures show

Officers initiated scans involving 257 adults and 14 children during the year, including two children under the age of 10, according to the released figures. More than half of the cases involved people suspected of an offence, being unlawfully at large or breaching bail conditions. A suspicion is not the same as a conviction.

The 24% share and the 7% population figure measure different things. The publicly reported account does not provide the full underlying dataset, including how any missing ethnicity information was handled or a detailed breakdown by age, area and reason for each check. Those details would matter when assessing the size and causes of the difference.

The figures describe who was selected for this type of search. They do not, on their own, measure whether the matching algorithm performs differently for particular groups. The decisions that lead to a person being photographed also deserve separate scrutiny from the technical performance of a system after an image is submitted.

Response and oversight

Civil liberties and equality campaigners have raised concerns that the use of biometric tools could reinforce existing inequalities. They argue that attention should be paid to who is stopped and scanned as well as to the risk of mistaken identification.

South Wales Police says its number of searches remains relatively small and that each use depends on the individual circumstances officers are dealing with. It says deployments are subject to supervisory and force-wide scrutiny, with independent advisory groups and external panels contributing to its governance. It has said it will take appropriate action where opportunities for improvement are identified.

The force also says independent testing found no identifiable bias on grounds of race, age or gender in the mobile system. That is a statement about the system’s performance in testing; it does not settle the separate question of how people come to be selected for a check.

The government has proposed new rules and independent oversight for police facial recognition. Those proposals are not a finding about the South Wales figures, nor do they mean a new law is already in force. Police use remains subject to existing requirements, including data protection, equality and human rights law.

No outcome of the force’s ongoing review has been announced in the reported response. The available figures do not establish a finding of unlawful discrimination or a decision to suspend use of the app.

Pin It on Pinterest