Amazon Web Services’ facial recognition service incorrectly matched 28 members of Congress to mugshots of people who have been arrested for crimes, the American Civil Liberties Union said in a blog post about an experiment conducted by the organization. The ACLU’s findings add to to the growing controversy over the use of facial recognition technology by law enforcement agencies.
Facial recognition technology is notoriously inaccurate in identifying people of color, and the same pattern was evident in the ACLU’s findings. The AWS service, called Rekognition, incorrectly matched six members of the Congressional Black Caucus, including Rep. John Lewis (D-Ga.), the ACLU said. The inaccuracies point to a larger challenge that all artificial intelligence companies face: Ensuring that the data used to train AI models doesn’t perpetuate discrimination against minorities and other groups.
An AWS spokeswoman, in a statement sent to media outlets, said the ACLU did not configure the Rekognition service to accurately identify people in its experiment.
Meanwhile, Democratic Sen. Ed Markey and two members of Congress wrote to Amazon CEO Jeff Bezos asking for more details. Given the scrutiny of tech power in Washington these days, it may only be a matter of time before Amazon, like Facebook, is called in to explain itself in a congressional hearing.