Facial recognition technology dating back 60 years advanced significantly with deep-learning methods over the past decade, enabling adoption by retailers, neighbors, and law enforcement according to IEEE Spectrum. The system must balance false positives and false negatives. Best-case passport scenarios show false-negative rates of two in 1,000 and false positives less than one in 1 million, per the report citing controlled tests.

False positives led to the 2020 wrongful arrest of Robert Williams by Detroit police, resulting in a settlement mandating policy changes on FRT limits, as documented in the IEEE Spectrum article and contemporaneous news reports. A 2023 federal court banned Rite Aid from using facial recognition for five years over a racially biased algorithm, per FTC records. UK government estimates indicated misidentification risks up to two orders of magnitude higher for women and darker-skinned people, cited directly in the Spectrum analysis.

U.S. Immigration and Customs Enforcement conducted more than 100,000 FRT searches via the Mobile Fortify app in its first six months since June 2025 against a gallery of at least 1.2 billion images, according to the IEEE Spectrum report. At 99.9 percent accuracy this scale implies around 1 million false matches, with elevated rates for specific subgroups. University of Massachusetts Amherst computer scientist Erik Learned-Miller stated that care in deploying such systems should be proportional to the stakes.