Always thought politicians were crooks? Now Amazon's facial recognition system agrees with you, but not in a good way. A test by the ACLU matching all 535 Congress members to 25,000 mugshots found 28 false positives—with a notably high number of people of color, the Verge reports. "An identification—whether accurate or not—could cost people their freedom or even their lives," says the ACLU. "Congress must take these threats seriously, hit the brakes, and enact a moratorium on law enforcement use of face recognition." But Amazon defended its Rekognition API software, saying the ACLU used the default certainty threshold of 80%, while Amazon suggests cops crank it up to at least 95%. Yet there's nothing to make law enforcement use the higher setting.
Rekognition drew attention in May when the ACLU revealed that several law enforcement agencies were using it. Privacy and civil rights groups cried foul, but police liked the price tag (around $12 a month), and at least one police force kept using it after a trial run, per USA Today. Some are concerned about racial bias: 39% of false positives in the ACLU test are people of color, including civil-rights leader John Lewis, while people of color comprise only 20% of Congress. Another test of facial-recognition software found higher false positives among blacks and women. Now three members of Congress have written Jeff Bezos expressing concerns about Rekognition. "Serious concerns have been raised about the dangers facial recognition can pose to privacy and civil rights," they write. (Read more Amazon.com stories.)