Salt Lake DA Drops $5,000 Felony Case After 94% AI Match Misidentified Utah Man
Updated
Updated · KUTV 2News · Jun 18
Salt Lake DA Drops $5,000 Felony Case After 94% AI Match Misidentified Utah Man
3 articles · Updated · KUTV 2News · Jun 18
Summary
Five months after Brad Johnston was charged with felony vandalism, Salt Lake County DA Sim Gill said the case will not be refiled because another culprit has been identified.
A facial-recognition search had linked Johnston to an Uber car vandalism case as a 94% match—97% in one police report—and prosecutors said it was presented as a positive identification.
Johnston's lawyer countered with bank records, photos and ride-history evidence she said showed he was not at the pickup bar that night, then pointed investigators to the person she says was the real suspect.
Salt Lake police still say that evidence does not exclude Johnston and cited the Uber video and his alleged association with another rider, while Gill called the case a lesson in using facial recognition only as a supplement.
Police AI has wrongly accused at least 14 people. Why do most states still lack laws to stop the next victim?
When an AI's 97% match is 100% wrong, what protects citizens from becoming ghosts in the justice system?
7 Wrongful Arrests and Counting: How Facial Recognition AI Is Failing Justice and What Must Change
Overview
Facial recognition AI works by finding a face in an image, creating a unique faceprint from key features, and comparing it to large databases often built from government photos. However, even a high match does not guarantee the person is correctly identified, leading to the 'best match paradox' where innocent people can be wrongly accused. These technical flaws, combined with systemic issues like bias and lack of transparency, have resulted in wrongful arrests and highlight the urgent need for strict safeguards, independent oversight, and policy reforms to protect civil liberties and public trust.