Randall Reid lives in Georgia. He’s 28 years old. He and his mother were headed to Thanksgiving dinner when the police arrived.

They accused Reid of stealing luxury purses from a shop in Louisiana. They had a warrant for his arrest. The warrant came from Jefferson Parish.

“What is Jefferson Parish?” Reid asked.

Jefferson Parish is an area of Louisiana, like a county. Reid would have known that—if he had ever been there. But he had never set foot in Louisiana. He had never stolen anything, either.

Thanksgiving dinner would have to wait. The police arrested Reid. He spent the next week in jail.

How did this mix-up happen? It’s all about facial recognition technology.

Machine learning (or “artificial intelligence”) can be trained to recognize specific human faces. Many people use this technology every day to unlock their cell phones. Facial recognition technology has boomed over the last decade.

As humans, we naturally recognize faces. You could easily pick your best friend out of a crowd. But for computers to imitate that skill, it takes complex math. Facial recognition uses measurements called “landmarks” to learn the human face. How wide apart are your eyes? What’s the distance between your nose and your mouth? These landmarks add up to a unique picture. With enough data, a computer can identify a face almost without fail.

Emphasis on almost.

Sometimes facial recognition makes mistakes. Studies suggest it makes even more mistakes when identifying people of color, like Reid. Even when the technology works, it raises concerns. What happens to privacy when governments and businesses can find your face in any photo or video?

Several U.S. states have passed laws to constrain the government’s use of facial recognition. In Louisiana, some legislators tried to pass such laws in 2021. The legislation didn’t pass.

That’s why Randall Reid spent the week after Thanksgiving in jail. A computer determined his face belonged to a thief three states away.

Thankfully for Reid, the Jefferson Parish sheriff noticed something fishy. Reid has a mole on his face; the thief did not. Reid appeared to weigh at least 40 pounds less than the thief. The sheriff dropped the warrant and released Reid.

According to the New Orleans police, facial recognition must go through department officials before leading to an arrest. But that didn’t stop an innocent man from being incarcerated.

Reid’s case raises questions about the government’s use of technology. Facial recognition might help keep the streets safe—for some. But that safety might come at the cost of privacy and freedom. Is it worth the trade?

Why? New technologies can open worlds of possibility. But if adopted without care, safeguards, and precision, they may cause more harm than good.