Arrested by an AI That Was Wrong

Our Take

Credit to the Unfair AI channel for laying out what happened to Nijeer Parks in plain terms: a facial recognition system got it wrong, and a man spent ten days in jail and lost his job because police trusted the algorithm's output as if it were fact. This isn't a hypothetical about future risk — it already happened, and it's still happening in departments across the country that treat AI matches as probable cause instead of a lead that needs real corroboration.

We cover ALPRs specifically, but the Parks case is part of the same story. Flock Safety cameras don't just read plates — they build location histories, feed 'convoy' and pattern analysis, and get shared across agencies with almost no independent audit of accuracy or misuse. Just like facial recognition, the selling point is speed and certainty, and the danger is that speed and certainty are often illusions. A wrong match, an outdated hotlist entry, or a typo in a plate reader database can put an innocent person's name in a police report the same way Parks' face did.

The fix isn't waiting for the tech to get better — it's refusing to let unaccountable systems make decisions about who gets stopped, detained, or arrested. Check our camera map to see what's watching your streets, and if you want to push back on how this surveillance gets deployed in your town, our take-action page has concrete next steps.