We might believe all politicians are crooks, but it transpired Amazon’s controversial facial recognition system actually matched US Congressional politicians to mugshots of suspected criminals in a study conducted by the American Civil Liberties Union (ACLU).
The ACLU has long called for the banning of facial recognition use by police forces as it has proven to be unreliable at best and dangerous at its worst.
The ACLU stepped-up its campaign after it was revealed Amazon had been aggressively marketing its AI-powered Rekognition Service to police forces across the US. It had been touted as a way to identify people in real time from surveillance camera footage of from officers’ own body cameras.
In its latest testing Rekognition mistook images of 28 Congressional members for mugshots of criminal suspects. ACLU found the incorrect matches were biased towards black people, including six members of the Congressional Black Caucus. In all, 40% of the wrong matches were of people of colour.
ACLU’s test involved comparing images of every senator and representative, all 535 of them, against 25,000 publicly available mugshots of arrested Americans. The 28 mismatches represented a 5% error rate.
While this might sound a reasonable error rate, applied to thousands of faces taken from security and body cameras, that could be extrapolated to a large number of innocent people being wrongly identified.
Couple this with automated fines and punishments then an innocent person could soon find themselves entering a Kafkaesque world of mistaken identity.
Facial recognition technology in the hands of government is primed for abuse and raises significant civil rights concerns. It could allow, and in some cases has already enabled, police to determine who attends protests, continuously monitor immigrants and cities to routinely track their own residents, whether they have reason to suspect criminal activity or not.- Jacob Snow: Attorney, ACLU
The ACLU has called for a moratorium to prohibit law enforcement agencies using facial recognition until further studies have been produced and Amazon has been more open about the system. Currently, it is not clear how the Rekognition system works particularly as Amazon has not disclosed how it tests for bias or accuracy.
Already a number of congress members have written to Amazon demanding a meeting to discuss in detail their Rekognition system.
In response, Amazon insisted that its system had been beneficial in preventing human trafficking, inhibiting child exploitation and reuniting missing people. With regard to ACLU’s latest survey, it said that at all times the system should be used in conjunction with human elements and acceptable thresholds should be set at 95% or higher.