AI Facial Recognition Failures: Bias, Bad Data, and False Arrests #shorts
The short version: Dave Linthicum's short breaks down how biased data and bad image quality make facial recognition misfire, a warning that applies just as much to the license-plate readers quietly tracking Americans' drivers.
Our Take
Dave Linthicum's short is a quick, useful primer on why facial recognition keeps getting it wrong: skewed training data, lousy image quality, and algorithms that essentially guess and call it a match. The false arrests that follow aren't edge cases — they're the predictable output of systems built on shaky foundations and sold to police as near-infallible.
We cover license plate readers, not faces, but the underlying problem is identical. Flock Safety's ALPR network runs on the same logic: trust the machine, skip the human verification, and scale it to every intersection in town. A blurry plate, a dirty camera lens, or a database error can turn an innocent driver into a felony stop just as easily as a bad facial match can turn a bystander into a suspect. Both systems convert probabilistic guesses into probable cause, and both put the burden of error on the public rather than the vendor.
The bigger issue is scale. One bad facial recognition hit is a tragedy for one family. A nationwide grid of ALPR cameras logging every car's movements is mass surveillance infrastructure, bias and all. Check our map to see what's already watching your neighborhood, and visit take action if you want to push back before the next 'false match' has your license plate attached to it.
This is DeFlock The USA’s original commentary. The video above is the work of Dave Linthicum Is Not AI, published on YouTube — full credit to the creator.