Facial Recognition Error: Man Suing Police After Wrong Arrest

Alvi Choudhury, 26, is pursuing legal action against Thames Valley Police and Hampshire Constabulary after being wrongfully arrested due to a facial recognition error. The software incorrectly matched him to a £3,000 burglary suspect in Milton Keynes, a town 100 miles from his home in Southampton where he was working. Mr. Choudhury was detained for 10 hours before being released with no further action.

He contends that CCTV footage of the alleged crime depicted a younger individual, significantly different from him in appearance, with the only similarity being curly hair. Mr. Choudhury's face was on the police database due to a prior wrongful arrest in 2021. The UK police forces utilize a German algorithm processing approximately 19 million mugshots, with searches conducted around 25,000 times monthly.

Despite the police stating a human assessment corroborated the technology's match, they admitted the mistake could stem from bias within the facial recognition technology. Research indicates higher false positive rates for Black (5.5%) and Asian (4%) faces compared to white faces (0.04%). Mr. Choudhury expressed concern that a second mugshot on the system could lead to future wrongful arrests, potentially impacting his software engineering career.

This case echoes similar incidents, including South Wales Police paying damages to a man wrongly identified as a stalking suspect, and an innocent grandfather mistaken for a shoplifter due to facial recognition. Police and crime commissioners have previously warned of inherent biases in such technologies, emphasizing that current lack of adverse impact is due to luck rather than design.