Falsely Arrested: Man Sues Police Over AI Error

0:00On March 18th, a 26-year-old software engineer named Alvi Chowdhury was arrested at his home in Southampton on January 7th after police. Facial recognition software wrongly matched his old mugshot to a burglar caught on camera in Milton Keynes. He spent 10 hours in custody before being released with no charges. The system flagged him for a theft at a Buddhist center where 3,000 pounds in jewelry were stolen from a charity fund. The match came from a mugshot taken 5 years earlier after another wrongful arrest during a street attack on Trowdhury and his friends. Meanwhile, police had already caught the real suspect, 23-year-old Edoord Zelatinyanu, who pleaded guilty just days later. A second thief remains at large, but officers relied on their own visual check after the AI alert before making the arrest. Chowdhury blames the technology's high-error rate, especially for Asian faces at 4 percent false matches, and what he calls lazy detective work, possibly tainted by racial bias.

1:04He's suing Tim's Valley Police for compensation and an apology, worried it could hurt his job with his security clearance. He points out no tech firm would release software this unreliable. Police insists the arrest followed a human review of footage, not just the AI, and apologize for the distress, but defend their tools as intelligence, leads. Home office studies show even worse false positives for black faces at 5.5% versus 0.04% for white faces. Other cases, like a man ejected from a store or damage is paid to a stalking suspect match, highlight ongoing flaws. Chowdhury now pushes for government oversight, new laws on AI use and policing, and better training to prevent repeats. With his new custody photo in the system, he fears more mix-ups ahead, unless changes come soon.