When facial recognition goes wrong

Our Take

ReasonTV's look at Robert Williams' case is a gut punch: a Michigan man spent a night in jail because a facial recognition match told cops he was their guy. He wasn't. The tech was wrong, and nobody caught it until real harm was already done. Credit to ReasonTV for keeping this story in front of people, because it's the clearest, most human example of what happens when law enforcement treats algorithmic output as ground truth instead of a lead that needs actual verification.

We cover license plate readers, not face scans, but the underlying rot is identical. Flock Safety and its ALPR network run on the same premise: a machine flags you, and that flag gets treated as probable cause, evidence, or justification for a stop. Nobody votes on whether these systems get deployed in their town, few departments publish accuracy audits, and when the tech gets it wrong, the burden of proving your innocence falls on you, not on the vendor or the agency that trusted the black box. Williams' story is what happens when that trust goes unchecked with a face. It's already happening with plates, every day, in thousands of jurisdictions with zero public oversight.

If you want to see how far this has spread near you, check our camera map, and if you want to push back on unaccountable surveillance infrastructure in your own community, our take-action page has concrete steps to start with.