When it comes to law enforcement’s involvement with the software, 63 percent of voters said they approve of such groups using the technology. However, African Americans, at 49 percent, were 16 points less likely than white voters (65 percent) to support law enforcement’s use of facial recognition, and Hispanics were 9 points less likely than white voters to say the same.
In July 2018, the American Civil Liberties Union released the results of a test of Amazon.com Inc.’s facial recognition tool, Rekognition, which a number of state law enforcement agencies use. In the test, the ACLU used Rekognition to see if it could correctly identify Washington lawmakers and found that the software misidentified 28 lawmakers — many of whom are racial minorities — as people who had been arrested for a crime. Amazon later said the ACLU was using the technology at a confidence threshold of 80 percent, when it was intended to be used at a 95 percent confidence level.
People of color are also less likely to support the use of facial recognition software by national security agencies: Sixty-seven percent of white voters support it, compared to 59 percent of African American voters and 58 percent of Hispanics.