AI Falsely Arrests Innocent Man, Raising Bias Concerns

0:00On April 30th, a 26-year-old software engineer named Alvi Shoudhury got wrongly arrested at his home in Southampton on January 7th because an AI facial recognition system from Tim's Valley Police matched him to a burglar caught on CCTV. The system pulled his old mugshot from a previous false arrest five years back and linked it to a theft at a Buddhist center in Milton Keynes, where 3,000 pounds in cash and some jewelry got stolen a month earlier. The real thief turned out to be Edaward Zalateanu, a 23-year-old from Birmingham, who got sentenced to 21 months in jail just five days after Chaudhury's arrest. Zalateanu was nabbed on the day of the crime, December 8th, but police still hauled in the innocent guy over four weeks later and held him for 10 hours, until they saw the footage at 2am and realized the mistake. Chaudhury and Zalateanu look nothing alike, one as curly hair, a thicker beard, thinner

1:01mustache and bigger eyebrows while the other sports short. Black hair. They live over 140 miles apart on opposite sides of the country, raising big questions about how officers visually confirmed the match after the AI flagged it. Chaudhury blames the text 4% false positive rate for Asian faces, plus what he calls sloppy detective work that might involve bias. He's suing the police for cash and an apology, worried this could mess with his job since he holds security clearances and notes his prior to. 1021 false arrests left his photo in the database despite promises to delete it. Home office studies show even higher error rates for black faces at 5.5%, compared to 0.04% for white ones. Fueling calls for better oversight on these tools that scan millions of muck shots monthly. Chaudhury wants new laws to hold everyone accountable before more innocence get caught in the net.