Independent studies, including from MIT and the federal government's own testing program, have found facial recognition systems produce higher error rates identifying women and people with darker skin tones. In one widely cited federal study, error rates for some systems were dozens of times higher for Black and Asian faces compared to white male faces, a gap advocates say has real consequences. Advocates argue error rates of this magnitude represent a serious, documented fairness problem, not a marginal technical issue. This is seen as a core reason for caution before wider deployment. They see error rates at this scale as a serious, not marginal, fairness concern.