AI Sent an Innocent Grandma to Jail for 5 Months

On July 14, 2025, United States Marshalls entered the home of 49-year-old Tennessee woman Angela Lipps and arrested her in front of four young children that she was babysitting. 

Authorities believed that Lipps had defrauded a series of North Dakota banks. That’s because the West Fargo Police Department had fed a photo ID used by the suspect into Clearview AI—a database with billions of pictures, many scraped from the internet, intended to aid law enforcement investigation. Lipps had been flagged as a match.

But the Tennessee grandmother had never even been to North Dakota. 

And so for five months, she sat in jail because of an AI-inspired wrongful arrest.

According to a GoFundMe set up in Lipp’s name (and verified by her attorney, Eric Rice), Lipps made multiple bank transactions that prove she was in Tennessee the entire time. She claims she was released from jail after five minutes of conversation with a detective explaining as much. 

In a written statement to A&E Crime + Investigation, interim Fargo Police Department Chief Travis Stefonowicz said that the department had “conducted a comprehensive internal review” and that “our initial investigation could have been more complete.” He further added that, in the time since, the department had adopted a formal policy around facial recognition technology. Stefonowicz stopped short of apologizing to Lipps or clearing her name, adding that “charges may be re-filed if additional investigation supports doing so."

Rice, who Lipps has retained as she pursues civil litigation in her wrongful arrest, tells A&E Crime + Investigation that the ordeal of detainment put a tremendous burden on her.

“She was transported halfway across the country. She’d never been to North Dakota, had no connections in North Dakota and didn’t know how long she would be detained,” he explains. “But her [financial] obligations, like housing and cost-of-living expenses, continued.”

On her GoFundMe, Lipps claims that during her lengthy detainment, she lost her rental home, car, dog, health insurance and social security income, amongst other things. She also claims she gained 65 lbs. during her time in jail. 

Rice says that he believes none of this would’ve happened had Lipps come from more wealth.

“People of lesser means get lesser justice,” Rice says. 

A spokeswoman for Clearview AI said in a statement to A&E Crime + Investigation, “Clearview AI's technology is designed to function as one tool within a broader investigative process. It generates leads, it does not make identifications, draw conclusions or recommend arrests.”

Still, some of Clearview AI’s critics say the tool’s flaws far outweigh its benefits—and they’re demanding a moratorium on its use. 

That’s the position held by Nate Freed Wessler, a civil rights attorney and the deputy director of the American Civil Liberties Union (ACLU)’s speech, privacy and technology project. 

"The problem here is that when facial recognition gets it wrong, it will get it wrong with a face that looks similar to the suspect,” Wessler tells A&E Crime + Investigation. “And that’s incredibly dangerous, because it taints the rest of the investigation.”

Wessler litigated the first wrongful arrest AI facial recognition case, of Michigan resident Robert Williams. At press time, Wessler says there are 14 known cases of wrongful arrest because of AI facial recognition technology, but he estimates that is a massive undercounting, because he believes that police departments are “often not informing criminal defense attorneys or even judges [that they’ve used the technology] when seeking these arrest warrants.” 

Different police departments use the technology at different scopes, with some restricting their searches to local mugshot databases and others to state driver’s license and identification databases. But the most dangerous, Wessler says, are the ones that use Clearview AI.

“They want every adult’s face on Earth in their database,” Wessler says of Clearview AI. “And each time a search is being run, we’re all being put into a lineup.” 

Both Wessler and Rice say that there are several safeguards that investigators could use to make this emerging technology better. 

Rice believes that the Fargo Police Department failed to run basic double-checking protocol such as putting Lipps’s photograph into a “six pack” of non-suspects and having eyewitnesses pick her out of a virtual lineup.

“We expect officers to do their investigative work,” he says. “This was an officer using AI as a shortcut.”

But Wessler says that even then the technology is dangerous, because people are innately biased to trust information provided to them by computer results—a phenomenon known as “automation bias.” 

Wessler also notes that AI facial recognition is less accurate at identifying some groups—including women.

Although he says the ACLU’s position is a “durable ban” nationwide on the technology in police work, he knows that some places don’t have the political will to make that a reality. In those areas, he believes, there need to be basic safeguards.

In Lipps’s bank fraud case, Wessler says, “Police had fingerprint and DNA evidence from the perpetrator. Maybe they should check the alibi. Maybe they should get a warrant for cellphone location history. If police were to get an anonymous tip, they’d do all kinds of things to corroborate or discorroborate it. And that’s what they should do here, because that’s about how reliable [this technology] is. Sometimes face recognition will get it right. But a lot of times, it’ll get it wrong.”