Massive errors found in facial recognition tech: US study

Facial recognition systems can produce wildly'inaccurate results', especially for Asian and African Americans, a new US government study has found. The new research comes amid widespread deployment of facial recognition technology for law enforcement, airports, banking, retailing and smartphones. The study found that facial recognition software would confused two people 100 times more often for Asian and African American faces than it did for white ones. Failures could lead to the'wrong people being arrested' and'lengthy interrogations' according to Jay Stanley of the American Civil Liberties Union. The National Institute of Standards and Technology (NIST) study also found two algorithms assigned the wrong gender to black females 35 per cent of the time.