Highlights
- Facial recognition misidentifies Southampton software engineer, leading to wrongful arrest.
- Technology produces higher false positives for south Asian and Black individuals.
- Home Office reviews algorithm and police guidance while calls for reform grow.
Choudhury, who wears a beard, told The Guardian the suspect looked “about 10 years younger,” with lighter skin, bigger nose, smaller lips, and no facial hair.
“Everything was different. I assumed the officer saw that I was brown with curly hair and decided to arrest me,” he said.
He was held in custody for nearly 10 hours before release at 2am, unable to work the following day, while his father was left anxious.
Thames Valley police defended the arrest, stating officers conducted a visual assessment after the facial recognition match.
The system, from German firm Cognitec, runs 25,000 searches monthly against 19 million UK police mugshots. The National Police Chiefs’ Council emphasises matches should guide investigations, not serve as evidence.
Bias raises alarm
Research commissioned by the Home Office revealed false positive rates of 5.5 per cent for Black faces and 4.0 per cent for Asian faces, compared with 0.04 per cent for white faces.
Choudhury, already in the system due to a prior mistaken arrest in 2021, worries further wrongful arrests may occur. He told The Guardian “In my head, if a brown person in Scotland robs a bank, are they going to come and arrest me?”
Live facial recognition has been deployed by Thames Valley police in Oxford, Slough, Reading, Wycombe, and Milton Keynes, capturing around 100,000 faces and leading to six arrests.
Civil rights experts, including Choudhury’s lawyer, stress AI must complement human judgment, not replace it.
The Home Office is reviewing guidance and training for retrospective facial recognition and developing a national system with an improved, independently tested algorithm.
Choudhury also raised concerns about security clearance for government work, noting repeated arrests can harm his reputation.
