SAN FRANCISCO (Reuters) – A face recognition tool like Amazon.com Inc. ( AMZN.O ) sells to web developers wrongly identified 28 members of the US Congress as police suspected in a test conducted by the American Civil Liberties Union (ACLU), the organization said on Thursday.
PHOTO PHOTO: Amazon.com logo is viewed at Amazon Japan's office building in Tokyo, Japan, August 8, 201
Amazon, in response, said that it addressed the face settings -ID tool during the test. However, the results highlight the risk that individuals can face if the police use the technology in certain ways to capture criminals.
Since May, ACLU and other civil rights groups have pushed Amazon to stop selling governments' access to Recognition, a powerful image ID software unveiled in 2016 by the company's cloud computing division.
The groups cited use of Law Enforcement in Oregon and Florida and warned that the tool could be used to target immigrants and colorants unfairly.
Their activism has kicked off a public debate. President of Microsoft Corp. MSFT.O ), Amazonian rival who also uses facial recognition technology, called at Congress earlier this month to study possible rules.
On Thursday, three Democrats identified in the ACLU test – Senator Edward Markey, Representative Luis Gutiérrez and Representative Mark DeSaulnier – sent a letter to Jeff Bezos, Amazon's Managing Director, who expressed concern and asked about the accuracy of the tool and the use of law enforcement .
Face Detection is already widely used in China for policy purposes, and a number of startup companies – some valued at billions – work aggressively with technology.
Amazon has asked a number of Recognition Uses, from detecting offensive content online to identify celebrities.
"We are still excited about how image and video analysis can be a driver for the world's best," said a spokeswoman for Amazon Web Services in a statement referring to help finding lost children and preventing crimes. She said that Recognition was usually used to limit the field of human review, not to make final decisions.
ACLUen said that it paid only $ 12.33 to have Amazon Recognition compare official images from each member of the US House and Senate against a database of 25,000 public arrest warrant.
The technology found "matches" for 28 members of the congress with at least 80 percent confidence, says Amazon's default setting, ACLU.
These struggles were disproportionate people of color, according to ACLU. About 39 percent were African-American and Latino legislators, against 20 percent identifying as a person of color in Congress, ACLU said. It is suggested that Recognition can exacerbate injury because people of color are already targeting an above average of the police.
"Facial monitoring is incorrect and it's preconceived and it's dangerous," said Jacob Snow, a lawyer at ACLU in northern California, Reuters.
Amazonian spokeswoman said the 80 percent confidence setting was suitable for identifying objects, not individuals. The company guides clients to set a threshold of 95 percent or higher for law enforcement activities, she said.
Reporting by Jeffrey Dastin in San Francisco; Editing Adrian Croft and Lisa Shumaker