False Rejections, True Identities: Facial Biometric Verification and the Structural Exclusion of Transgender Users in Brazil's Digital Identity Infrastructure

Over 174 million Brazilians now hold an identity account with the federal government's Gov.br platform, and roughly 80 million have reached its highest trust tier by matching a live selfie against a photograph already on file with the electoral court, the traffic authority, or the national ID system. Separately, three in four Brazilian banks now identify customers with facial biometrics, often as a direct replacement for fingerprint sensors that had their own well-documented failure mode: worn, chemically damaged, or aged fingerprints among manual laborers and the elderly. That substitution fixed one exclusion and quietly opened another. A fingerprint does not change when someone transitions gender; a face does, and it keeps not matching long after the transition itself is finished, not only during it. This paper's central concern is that population: transgender people whose facial characteristics changed as a result of transition, in ways that continue to affect facematch outcomes long after transition is complete. We trace how Gov.br's face-matching pipeline decides who gets to be recognized as themselves, situate this within Brazil's own regulatory debate over whether biometric enrollment can really be refused, and compare it to two other countries (India and the United Kingdom) where biometric identity infrastructure has produced comparable, independently documented exclusion of vulnerable citizens. We call this the transition gap, and argue that trans people deserve to be part of how a facematch system is designed and trained from the start, not a correction applied after deployment once fairness metrics happen to surface a gap.