ACLU calls out Amazon biometrics in fight to ban facial recognition
The American Civil Liberties Union of Northern California (ACLU) has used Amazon Rekognition facial recognition to match one in five California lawmakers with arrest mugshots, using Rekognition’s default settings, including an 80 percent confidence threshold, Vice reports.
At a press conference held by the ACLU and legislators, ACLU attorney Matt Cagle said 26 state legislators were matched out of 120 compared to a database of 25,000 arrest photos. Assemblymember Reggie Jones-Sawyer said that facial recognition would automate mistaken identity and entrench racial bias in policing.
Improving technology and meaningful oversight could enable facial recognition to be used while mitigating risk, Max Pellegrini, president of RealNetworks, which makes SAFR facial recognition products, told Biometric Update in an email.
“Today’s facial recognition systems are imperfect, demonstrating some level of inherent bias. However, facial recognition technology is rapidly improving and will eventually reflect less bias than humans,” argues Pellegrini. “Rather than banning the use of this valuable new technology, we should advocate oversight in the form of meaningful human review prior to making final decisions that result in a legal or other significant consequences for citizens. In concert with transparency, human oversight and third-party validation testing provide meaningful protection against unbiased results.”
Even if the algorithms were completely accurate, Cagle argues the technology would affect vulnerable communities more than others, and the bill notes some people could be dissuaded from talking to the police.
“This is not the sort of problem that can be solved simply by tweaking an algorithm,” Cagle said. “This is the kind of problem that the legislature needs to step up and fix right now to protect all Californians.”
Amazon recommends police use a confidence threshold of 99 percent, but one of the few known customers of the technology, Washington County Sheriff’s Office, has admitted it does not utilize a confidence threshold at all. A year ago, ACLU matched 28 members of U.S. Congress with mugshots from a database, which appears to be the same one.
The experiment by ACLU comes as legislation has been introduced in the state to block the use of facial recognition in police body cameras. The bill is co-sponsored by the ACLU and state Assemblymember Phil Ting, who was among the 26 misidentified lawmakers.
Ting noted at the press conference that the even an individual eventually cleared of wrongdoing can suffer consequences from a false arrest.
“It could impact your ability to get employment, it absolutely impacts your ability to get housing. There are real people who could have real impacts,” Ting said.