When a Lead Becomes Probable Cause: Facial Recognition, Investigative Escalation, and Mistaken Arrest

Technical error, false matches, and uneven model performance are important features of facial-recognition systems, but they are not the principal focus of this paper. This paper instead examines the institutional and procedural danger that arises when a tool introduced to generate investigative leads is treated as though it had already established identity. In that setting, a tentative return can be elevated into actionable suspicion and used to support a warrant request before the underlying basis for action has been adequately tested (DOJ, 2024, p.

23; Garvie, 2022). The resulting failure is therefore not merely a mistaken output. It is a constitutional failure with civil-rights consequences, including unlawful arrest, wrongful detention, and the use of the state’s arrest and detention power on inadequately tested grounds (Manuel v.

City of Joliet, 2017; DOJ, 2024, p. 11).