Cops Lost His Kids Over an 85% Guess — Your Face Could Be Next

A man lost custody of his two kids. He lost his job. He lost his home. And he spent two months locked up — all because a computer said his face was an eighty-five percent match to someone in a surveillance video. His alibi? He was hundreds of miles away when the crime happened.

Here's why this matters to you, even if you've

Here's why this matters to you, even if you've never been near a police station. Your face is probably already sitting in a database somewhere — a driver's license photo, a mugshot system, a social media scrape. You didn't sign up for it. According to the American Civil Liberties Union, at least fourteen people in the U.S. have been wrongfully arrested after a facial recognition system pointed at the wrong person. Most of them were Black. So the question threading through this whole story is simple. Why are police arresting people based on a probability — and treating a guess like a fact?

Let's start with that eighty-five percent number, because it sounds convincing. It isn't proof. It's a similarity score — a measure of how much two images look alike to software. In the recent case reported by Futurism, police took that match and paired it with two eyewitness accounts to claim they had probable cause. They had enough to arrest. But the man had a verified alibi, sitting hundreds of miles away. Nobody checked his location before they upended his life.

Now, why would the software get it wrong in the first place? Researchers tested one hundred eighty-nine different algorithms from nearly a hundred developers. Their finding, documented in analysis of N.I.S.T. data — that's the National Institute of Standards and Technology — was stark. Many of these systems misidentify Black and East Asian faces ten to a hundred times more often than white faces. Sit with that. A tool that's far more likely to be wrong about certain faces is being used to put those exact faces in handcuffs. For you, that means the system's reliability can change depending on what you look like.

And the people who build and oversee these tools know it. The Michigan State Police and the facial recognition vendor both say plainly — the results should never be the only reason to arrest someone. The instructions are right there. A face match is a lead, a starting point, not a conclusion. Yet case after case shows officers treating that score like a fingerprint. The difference matters. A fingerprint is near-certain. A similarity score is a maybe.