Wrong Face, Real Arrest: When Facial Recognition Fails
The short version: techwatchreport's video on facial recognition misidentifications is a reminder that automated 'matches' are guesses dressed up as evidence, and Flock's ALPR network runs on the same flawed logic.
Our Take
Credit to techwatchreport for laying out the human cost when facial recognition gets it wrong: real people, real arrests, real trauma, based on an algorithm's best guess. These cases keep surfacing because the technology is being trusted far beyond what its accuracy actually supports, and the people harmed are disproportionately Black and Brown residents who were never in the system's training data the way it should have accounted for.
We cover license plate readers, not face scans, but the underlying problem is identical. Flock Safety's ALPR cameras generate 'hits' on plates and vehicle descriptions that police treat as probable cause, even though the system is guessing at partial plates, similar makes, and low-resolution images. A wrong face leads to a wrongful arrest. A wrong plate read can do the exact same thing, and there's no independent audit forcing local police departments to prove otherwise.
The lesson isn't just about facial recognition, it's about what happens when we let machine outputs substitute for actual investigation. If your city is quietly running one of these camera networks, you deserve to know it. Check our camera map to see what's been deployed near you, and if you want to push back, our take-action page has concrete steps for getting these systems in front of your city council.
This is DeFlock The USA’s original commentary. The video above is the work of techwatchreport, published on YouTube — full credit to the creator.