Facial recognition uses AI to map facial geometry — the distance between eyes, nose shape, jawline — and match it against stored images. The technology has spread rapidly through law enforcement, airports, and consumer devices. A landmark NIST study found that many commercial facial recognition algorithms had error rates 10 to 100 times higher for Black and Asian faces than for white faces.
Facial recognition errors are not random — NIST tests show the technology misidentifies Black women at error rates up to 100 times higher than white men. When police, border agents, or employers use the technology to make consequential decisions, the people most likely to be wrongly identified are already the most vulnerable. Several cities have banned government use; Congress hasn't acted.
People assume facial recognition just matches a photo to an ID — a neutral process like checking a driver's license. Law enforcement use goes further: it compares faces against millions of photos in databases that include arrest records, DMV files, and social media scrapes, with no requirement that the match be disclosed to the person identified.
Facial recognition errors are not random — NIST tests show the technology misidentifies Black women at error rates up to 100 times higher than white men. When police, border agents, or employers use the technology to make consequential decisions, the people most likely to be wrongly identified are already the most vulnerable. Several cities have banned government use; Congress hasn't acted.
People assume facial recognition just matches a photo to an ID — a neutral process like checking a driver's license. Law enforcement use goes further: it compares faces against millions of photos in databases that include arrest records, DMV files, and social media scrapes, with no requirement that the match be disclosed to the person identified.