
South Wales Police data from 2025 shows that 24% of people selected for facial recognition checks were from ethnic minority backgrounds, despite this demographic comprising only 7% of the region’s population. The disparity prompted civil liberties groups and anti-discrimination organizations to voice concerns about potential racial bias in the deployment of the technology.
The police force, which serves Wales’s largest cities Cardiff and Swansea, uses a smartphone-based system called “Neoface” that captures facial images and matches them against custody databases and other face records. Officers initiated 257 scans of adults and 14 of children throughout 2025, primarily targeting individuals suspected of crimes, those unlawfully at large, or people posing perceived risks to themselves or others. Roughly half of the scans resulted in either no match or voluntary identification, with arrests following in 48% of cases where matches occurred.
Critics, including Big Brother Watch and Race Equality First, characterized the findings as alarming and indicative of systemic inequality. Concerns center on whether artificial intelligence systems adequately represent ethnic minority faces in their training data, potentially increasing false positive rates for non-white individuals. One man arrested following a facial recognition match for a burglary in an unfamiliar city expressed worry that the technology was unreliable and subject to operator bias.
South Wales Police acknowledged the data raised disproportionality concerns and committed to reviewing its operational approach through established governance processes. The force emphasized that each deployment decision reflects individual circumstances and undergoes supervisory oversight. The Home Office recently established a facial recognition watchdog with authority to halt police use of the technology for non-compliance with equality, human rights, and data protection requirements, while emphasizing that operational decisions remain with police forces provided they meet legal standards.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI