AI Jailed An Innocent Grandmother For 6 Months
The short version: A grandmother's six-month wrongful jailing shows what happens when machines accuse and no human bothers to double-check the work.
Our Take
Credit to the channel not SENTIENT yet for putting a human face on what usually gets buried in a press release: a facial recognition match, treated as gospel, that cost an innocent woman half a year of her life. This is the pattern we keep seeing with automated surveillance tech — a black box spits out a name, and the humans in the loop rubber-stamp it because pushing back takes effort nobody's paid to spend.
Flock Safety's ALPR network runs on the same logic. Cameras flag a plate as "suspicious," cops treat the algorithm's confidence as their own, and the burden of proof quietly shifts onto whoever got flagged. No independent audit, no meaningful appeal before the damage is done. Wrongful stops and wrongful arrests tied to plate-reader "hits" aren't hypothetical — they're the predictable output of a system built for speed, not accuracy.
The lesson isn't just about facial recognition — it's about what happens any time we let a machine's guess stand in for actual verification. Check our camera map to see how much of this infrastructure has quietly gone up in your area, and visit take action if you want to push back before your town's next "hit" is a person, not a plate.
This is DeFlock The USA’s original commentary. The video above is the work of not SENTIENT yet, published on YouTube — full credit to the creator.