Clearview AI wins biometrics contract with U.S. Immigration and Customs Enforcement amidst ongoing controversy
An investigation by industry watchdog Tech Inquiry has revealed that the U.S. Immigration and Customs Enforcement (ICE) agency paid $224,000 to Clearview AI for licenses to biometric facial recognition services on August 12, 2020.
Clearview CEO Hoan Ton-That confirmed the contract, though says it is with Homeland Security Investigations (HSI) in an email to The Verge, saying the technology is used by ICE’s “Child Exploitation Unit and ongoing criminal investigations. Ton-That claims the company’s technology has enabled the unit to rescue children from sexual abuse. The contracting office named in a document obtained by Tech Inquiry is “DHS: ICE; mission support dallas [sic].”
Federal records indicate Clearview entered into a $50,000 contract with the U.S. Air Force last December.
As it gears up to fight them, Clearview has retained Floyd Abrams, an attorney well-known for fighting first amendment cases, such as the Pentagon Papers, in which he represented the New York Times in 1971, and Citizens United, which preceded a flood of corporate money into political campaigns.
The Times reports that he has argued 13 cases before the Supreme Court. Abrams suggests that the First Amendment directly clashes with privacy claims against Clearview, and that he is unfamiliar with the technology.
Clearview claims to have a high-quality algorithm that is accurate for all demographics, including women and people with darker skin, but without offering supporting information.
MIT Technology Review points out in an article that there are now suspected to have been two different recent cases in the U.S. of police misconduct leading a lawful arrest being related to and blamed on biometric facial recognition systems, both of Black men. Assessing the claims of Clearview is impossible, as the Review notes, because the company has not submitted any algorithm for evaluation by the National Institute of Standards and Technology (NIST).
The article also repeats idiosyncratic interpretations of accuracy rates, and makes the puzzling claim, in the same paragraph as the NIST reference, that there is no independent test specifically for bias.