Police used AI facial recognition to arrest a Tennessee woman for crimes committed in a state she says she’s never visited

A Tennessee grandmother spent more than five months behind bars after police relied in part on an artificial intelligence facial recognition system that linked her to fraud crimes in North Dakota — a state she says she had never visited.

Authorities in Fargo later acknowledged mistakes in the investigation and said they are changing procedures, though they have not issued a direct apology.

Angela Lipps, 50, was arrested in Tennessee on July 14, according to the Fargo Police Department and a verified GoFundMe created on her behalf. At the time of her arrest, Lipps did not know that a warrant had been issued for her weeks earlier in Fargo, more than 1,000 miles away.

The warrant was connected to a series of bank fraud incidents that had taken place in the Fargo area. During their search for a suspect, investigators used facial recognition technology obtained through another law enforcement agency, along with additional investigative steps, before submitting the case to the Cass County State Attorney’s Office, Fargo Police Chief Dave Zibolski said in an email.

At a news conference Tuesday, Zibolski said that part of the problem stemmed from information provided by a neighboring department’s AI system. According to the chief, West Fargo police had purchased their own facial recognition program without Fargo leadership’s knowledge, and Fargo authorities would not have approved its use had they been aware of it. The technology has since been barred from use by Fargo investigators.

West Fargo police said their department uses Clearview AI, a company that compiles billions of images gathered from across the internet and social media platforms. The software flagged someone with facial features resembling Lipps as a possible suspect. That information was passed to Fargo police.

West Fargo officials noted they never filed charges in their own fraud investigation because they lacked sufficient evidence. It remains unclear what additional evidence was used to connect Lipps to the Fargo crimes.

The situation highlights growing concerns about law enforcement’s rapid adoption of artificial intelligence tools, particularly facial recognition systems that critics say can produce incorrect matches.

A judge in North Dakota approved a nationwide arrest warrant for Lipps on July 1. Two weeks later, she was taken into custody in Tennessee. She remained in a Tennessee jail for more than three months before being extradited to North Dakota.

It was not until October that Tennessee authorities notified the Cass County Sheriff’s Office that Lipps had agreed to extradition. She faced several charges, including felony theft and felony misuse of personal identification information.

Lipps’ attorneys told CNN they had seen a July 14 email notifying North Dakota law enforcement officials of her arrest. Fargo police, however, said they could not determine whether the delay in transferring Lipps was caused by a probation violation in Tennessee or because she initially contested extradition.

Lipps described the experience of being transported to North Dakota as frightening. In her GoFundMe statement, she wrote that the flight was her first time ever on an airplane and said she felt “terrified and exhausted and humiliated.”

After arriving in Fargo, a defense attorney obtained bank records showing that Lipps had been in Tennessee when the fraudulent transactions occurred. According to police, the Cass County State Attorney’s Office was informed on December 12 that the defense had produced evidence suggesting she was not involved.

On December 23, the detective assigned to the case, prosecutors, and a judge agreed to dismiss the charges without prejudice so investigators could continue reviewing the evidence. Lipps was released from custody on Christmas Eve.

The months she spent jailed had serious consequences, her lawyers said.

“The trauma, loss of liberty, and reputational damage cannot be easily fixed,” they said in an email, noting that Lipps was not available for interviews.

Lipps is a mother of three and grandmother of five. According to local media outlet WDAY, she had never traveled to North Dakota before being extradited there. After her experience, she said she does not intend to return to the state.

Her legal team is examining why she remained incarcerated for so long despite records showing she was elsewhere during the crimes.

“We believe that Angela’s lengthy detention was unnecessary and should have been avoided with a proper investigation by law enforcement,” the attorneys said. They are considering civil rights claims but have not yet filed a lawsuit.

During the Tuesday news conference, Chief Zibolski said investigators identified “a couple of errors” that contributed to Lipps being named as a suspect.

Although Fargo police do not operate their own AI facial recognition system, West Fargo does. The system flagged Lipps as a possible match to the photo on a fraudulent ID connected to one of the cases. According to Zibolski, Fargo detectives incorrectly assumed that surveillance images from the fraud incidents had also been submitted with the ID photo.

As a result of the case, Fargo police will stop using information generated by West Fargo’s AI system. Instead, they plan to work with state and federal partners, including the North Dakota State and Local Intelligence Center. The department will also begin reviewing all facial recognition identifications monthly to monitor how the technology is being used.

Zibolski said another mistake involved failing to send surveillance images from the fraud investigations to the state intelligence center, which is certified to conduct facial recognition analysis. After reviewing the footage, that center has identified other possible suspects.

The chief also addressed the delay between Lipps’ extradition and when Fargo investigators first spoke with her. He said there is currently no straightforward system that alerts Fargo police when someone arrested on one of their felony warrants is taken into custody elsewhere. The department is considering changes, including reviewing jail booking records daily.

When asked whether police planned to apologize to Lipps, Zibolski said investigators are still trying to determine who was involved in the fraud scheme.

“We still don’t know who’s involved and who’s not involved,” he said, adding that disciplinary action against officers involved in the investigation is still under consideration.

He said the officers connected to the case are frustrated by the outcome because they take pride in thorough investigations and do not want innocent people arrested.

The Cass County State Attorney’s Office is also exploring additional training on facial recognition technology through the North Dakota State and Local Intelligence Center so prosecutors can better understand how the systems work.

Fargo police previously said the investigation remains open and that charges could be filed again if further evidence supports doing so.

Lipps’ attorneys said they support improvements in how AI technology is used but criticized investigators for failing to conduct basic checks before seeking an arrest warrant.

They said officers were aware that Lipps lived in Tennessee and should have determined whether she was even in North Dakota when the crimes occurred.

“Instead, an officer used AI facial recognition as a shortcut for basic investigation,” the lawyers said in a statement, “resulting in an innocent woman being detained and transported halfway across the country to answer for charges that she had nothing to do with.”

The case is part of a broader debate over the growing use of artificial intelligence in policing.

In another incident last year, armed officers in Baltimore County detained and searched a high school student after an AI-based security system mistakenly flagged a bag of Doritos as a possible firearm.

The mistake led to criticism of the school’s security measures and calls for better oversight of automated systems.

Ian Adams, an assistant professor of criminology and criminal justice at the University of South Carolina, said law enforcement agencies are adopting AI technologies quickly, often without strong evidence of their reliability.

He said mistakes involving AI usually stem from a combination of technological limitations and human decisions.

“When technology is used incorrectly or without proper oversight, that’s when problems arise,” Adams said. He added that investigators must carefully review AI-generated results rather than relying on them without verification.