Amazon Rekognition wrongly matches 28 Members of Congress 

Occurred: July 2018

Can you improve this page?
Share your insights with us

Amazon's Rekognition facial recognition system incorrectly identified 28 members of Congress as other people who had been arrested for a criminal offences.

The American Civil Liberties Union (ACLU) released the results of a test showing that Rekognition had falsely matched 28 members of US Congress with mugshot photos of criminals, especially people of colour. Congressional members from both major political parties later expressed concern about Rekognition in a series (pdf) of letters to Amazon CEO Jeff Bezos.

Amazon responded by saying the Rekognition test had generated 80 percent confidence, and that it recommended law enforcement only use matches rated at 95 percent confidence or higher. According to the ACLU, Amazon moved the goalpost by increasing the recommended confidence interval after the ACLU study was published.

The incident raised questions about the accuracy and reliability of Amazon's Rekognition system and highlighted its potential impact on minority community civil liberties when used by law enforcement authorities.


Operator: American Civil Liberties Union (ACLU)
Developer: Amazon/AWS
Country: USA
Sector: Politics
Purpose: Identify public figures
Technology: Facial recognition
Issue: Accuracy/reliability; Bias/discrimination - race; Human/civil rights; Privacy
Transparency: Governance