Facial Recognition Error in Richardson Arrest Reveals Persistent Law Enforcement Verification Failures

Jacksonville Sheriff’s Office used AI facial recognition matching surveillance footage and a fake ID at 85 percent confidence to issue a warrant for Jalil Richardson in a 2025 stolen-vehicle case, resulting in his 50-day incarceration across two states before charges were dropped upon presentation of work time sheets proving he was 400 miles away. Court records and the WSOC-TV report confirm the arrest warrant relied on the AI output plus a subsequent lineup identification, with no pre-warrant verification of Richardson’s location or travel history despite the technology’s documented demographic performance gaps reported in NIST IR 8280. Similar outcomes appear in the 2020 Robert Williams arrest by Detroit police, where a faulty match from the same class of algorithms produced probable cause without corroboration, and in multiple cases tracked by the Georgetown Law Center on Privacy & Technology, indicating that treating facial recognition as one investigative tool routinely bypasses required independent verification steps.