Facials Fargo: Wrong Facial Match Jails Grandma 108 Days

Angela Lipps was babysitting four young children at home in Tennessee when U.S. Marshals arrived with their guns drawn. They arrested her for bank theft in Fargo, North Dakota. That's roughly twelve hundred miles from where she was standing. She spent a hundred and eight days in jail, and it started with a facial recognition match.

If there's a photo of you anywhere, from a driver's

If there's a photo of you anywhere, from a driver's license to a social media post, your face can be compared against a stranger's caught on camera. Most of us never think about that. Angela Lipps probably didn't either. According to Gadget Review, she's now suing the city of Fargo for ten million dollars. Her lawsuit says police treated a computer's guess as if it were proof. Fargo officials say they did more checking than that. In the end, detective work didn't free her. Bank records did. So the question running through this whole story is simple. What evidence did police actually have, besides the algorithm?

Start with the security footage itself. According to the lawsuit, the woman on the bank cameras didn't look much like Lipps up close. Her body type was different. Her facial features were different. Her tattoos were different. And investigators had evidence placing Lipps more than a thousand miles away when the thefts happened. She stayed locked up anyway. She got out on Christmas Eve.

A hundred and eight days. For a face that wasn't hers.

Fargo's police chief, Dave Zibolski, has admitted his department made what he called "a couple of errors." He also made a point about the fake I.D. at the center of the case. Just because a photo sits on a fake I.D., he said, doesn't mean the person in that photo is the one who used it. His department still pushed to have Lipps sent from Tennessee to North Dakota. If the chief saw the weak spot, why didn't the case slow down?

Part of the answer is how these systems really perform

Part of the answer is how these systems really perform. With crisp, well-lit photos, the best facial recognition programs get it wrong about once in a thousand tries. Real life isn't crisp and well-lit. With grainy, real-world images like bank camera stills, the error rate can climb to about one in five. That's a huge gap. For an investigator, it means a match from blurry footage is a starting point, not a finish line. For everyone else, it means the camera that caught a stranger could point at you instead.

And Lipps isn't alone. The American Civil Liberties Union has documented at least ten wrongful arrests tied to facial recognition, across eight states, through twenty twenty-four. The A.C.L.U. says that's a floor, not a ceiling. It only counts cases that made it to court or into the news. Research has also found these systems make more mistakes with women and with Black people. Black women are hit hardest. Nearly every known wrongful arrest in the U.S. from a bad face match has involved a Black person.

Some cities have already changed the rules. Detroit police can no longer arrest someone based only on a facial recognition result. They also can't run a photo lineup straight off a face match. Officers need separate, reliable evidence connecting that person to the crime. That's a written policy. For families, it's the difference between a knock on the door and a phone call asking where you were that day.

Fargo says its investigators did corroborate the lead. Lipps's lawyers say those steps were thin or never happened. A federal court will sort that out. For the lawyers and investigators listening, the case may turn less on the software's accuracy and more on due process. In other words, did police have real evidence before they acted?