When Face Recognition Doesn’t Know Your Face Is a Face (www.wired.com)

Autumn Gardiner’s simple DMV photo update turned into a humiliating ordeal when the office’s camera repeatedly rejected her—she lives with Freeman‑Sheldon syndrome, a facial difference that alters the appearance of facial muscles. WIRED’s reporting finds dozens of similar cases: people with craniofacial conditions, birthmarks or other facial differences being blocked by biometric face‑verification systems used by governments, banks, credit agencies, social platforms and phone makers. Face Equality International estimates more than 100 million people worldwide could be affected; failures range from being unable to open online Social Security accounts or retrieve credit scores to problems at passport control and airport gates. Technically, many of these systems create “faceprints” (measuring distances and landmarks like eye spacing or jaw contours) and run liveness checks, but they often rely on training datasets that under‑represent atypical faces, producing false negatives and no clear human fallback. The result is systemic exclusion amplified by automation: repeated rejections, boilerplate customer support, and inconsistent accommodations. The story underscores immediate fixes—mandated alternative authentication methods, better data collection and inclusive model training, human override protocols and industry audits—and longer‑term needs for regulation and collaboration with facial‑difference communities to make biometric systems reliable and equitable.