Detroit: Face recognition is almost always wrong

Detroit: Face recognition is almost always wrong

Detroit, 1.7.2020

Detroit police face recognition technology incorrectly identifies people in about 96% of cases, according to an estimate by Chief of Police James Craig. Craig admitted errors in the technology during a public session on Monday.

His statement came after the ACLU filed a complaint against the department for wrongly arresting a man for facial recognition errors.

Craig said that if the department were to use facial recognition software only to identify suspects, the cases would not be solved in 95 to 97% of cases.

The ACLU has called on Detroit and other police departments to stop using facial recognition and has called the technology „flawed.

Recently, the union filed a complaint on behalf of Robert Julian-Borchak Williams, a black man arrested by Detroit police after a facial recognition service falsely attributed him to a shoplifter. A video shows Detroit officials releasing Williams after they admitted that „the computer“ must have been wrong.

The department’s statistics show that this year police have used facial recognition technology almost exclusively on blacks, Vice reports.

DataWorks Plus, which provides Detroit’s facial recognition software, said there are no statistics on its accuracy.