Could Facial Recognition Get You Wrongly Arrested?

Our Take

Proton's report on Shaun Thompson isn't really a story about one bad camera in London — it's a warning about what happens when law enforcement outsources judgment to algorithms. A machine flags a face, a cop trusts the flag, and a man ends up detained for a crime he didn't commit. Swap "face" for "license plate" and you've got the exact failure mode we track every day here in the States with Flock Safety's ALPR network: a system that scans, matches, and alerts with a confidence it hasn't earned, while the human in the loop treats the algorithm's guess as ground truth.

Flock will tell you plates aren't faces, that this is just fleet-management software, nothing to see here. But the underlying problem is identical: these systems generate false positives, departments don't publish real error rates, and the burden of the mistake lands entirely on the person misidentified — not on the vendor selling the software or the officer who acted on it. Thompson's case shows how thin the line is between "investigative lead" and "probable cause" once a computer says a name with enough confidence. That's exactly the dynamic critics warn about when Flock hits are used to justify stops, searches, or worse, with zero transparency about how often the system gets it wrong.

Credit to Proton for putting a human face on what's usually discussed in the abstract. If you want to see how much of this infrastructure is already watching your own neighborhood, check our camera map, and if you're ready to push back on it locally, our take-action page is a good place to start.