
""A 2024 test by the National Institute of Standards and Technology found that facial recognition tools are less accurate when images are low quality, blurry, obscured, or taken from the side or in poor light-exactly the kind of images an ICE agent would likely capture when using a smartphone in the field," their letter said. If ICE's use continues to expand, mistakes "will almost certainly proliferate," senators said, and "even if ICE's facial recognition tools were perfectly accurate, these technologies would still pose serious threats to individual privacy and free speech.""
"Matthew Guariglia, senior policy analyst at the Electronic Frontier Foundation, told 404 Media that ICE's growing use of facial recognition confirms that "we should have banned government use of face recognition when we had the chance because it is dangerous, invasive, and an inherent threat to civil liberties." It also suggests that "any remaining pretense that ICE is harassing and surveilling people in any kind of 'precise' way should be left in the dust," Guariglia said. In their letter to ICE acting director Todd Lyons, senators sent a long list of questions to learn more about "ICE's expanded use of biometric technology systems," which senators suggested risked having "a sweeping and lasting impact on the public's civil rights and liberties." They demanded to know when ICE started using face scans in domestic deployments, as previously the technology was only known to be used at the border, and what testing was done to ensure apps like Mobile Fortify are accurate and unbiased."
A 2024 NIST test found facial recognition tools are less accurate with low-quality, blurry, obscured, angled, or poorly lit images, which mirror smartphone field captures. Expanded ICE use of facial recognition will likely increase mistakes and could produce sweeping, lasting impacts on civil rights and liberties. Even perfect accuracy would still pose serious threats to individual privacy and free speech. Questions were sent to ICE acting director Todd Lyons about domestic face-scan deployments, testing and bias mitigation for Mobile Fortify, and policies for identifying U.S. citizens; a requested response by October 2 was not received.
 Read at Ars Technica
Unable to calculate read time
 Collection 
[
|
 ... 
]