A Tennessee grandmother was babysitting her neighbor’s four children when armed officers came to arrest her.
It happened in July 2025. She was taken at gunpoint and spent months in jail. Now she is suing, saying an AI facial recognition result helped send her into a case that had nothing to do with her.
Her name is Angela Lipps. The lawsuit says she was caring for the children in Tennessee when officers arrived with a warrant tied to bank thefts in North Dakota. The charges were later dropped after records showed she was in Tennessee when the crimes happened.
That is the part people cannot shake.
The arrest was not a warning on a screen. It was not a wrong name in a database. It was a real arrest, with guns, jail, missed time with family, and months lost before anyone corrected the mistake.
Lipps is suing the city of Fargo and a Fargo police detective. She seeks $10 million. The complaint says officers treated an AI facial recognition result as far stronger than it was.
That claim is disputed in court. It is not a final ruling. But the public record already shows a basic failure: a tool meant to produce a lead became part of the path to an arrest.
The lawsuit says a West Fargo investigator used Clearview AI to search a photo from a fake ID. The search returned a social media photo of Lipps as a possible match. Clearview says its results should be treated as leads and checked through other methods.
The complaint says those checks should have stopped the case.
Investigators allegedly had surveillance images that did not match Lipps in build, facial features, and tattoos. The lawsuit also says she had no link to the bank thefts and was in Tennessee at the time. West Fargo police reportedly did not seek charges based on the facial recognition result alone.
The Fargo investigation then moved in another direction, according to the complaint. It alleges that Detective Lucas Heck relied on the result when seeking an arrest warrant and did not include key facts that pointed away from Lipps.
That is where the word “AI” can become too convenient.
The software may have produced a bad match. But software did not request the warrant. Software did not decide that a possible lead was enough. Software did not keep a person in jail for months.
People did those things.
The public reaction has split along familiar lines. Some of the anger is aimed at facial recognition. Some is aimed at police work that failed to check the result. Some of the reaction turns the story into a warning about AI itself. The jokes are easy to find in any technology scandal, but this story has made the room quieter than usual.
The quiet makes sense.
A person can avoid an app. They can delete a social media account. They cannot easily avoid the power of a police system that turns an image into suspicion, then turns suspicion into a warrant.
The facts still need care. The lawsuit alleges that AI was misused. The case will have to decide what each department knew, what each officer reviewed, and what supported the arrest. The public record does not prove every claim in the complaint.
The city of Fargo has said it does not comment on pending litigation. Fargo’s police chief later acknowledged investigative errors and said the department misunderstood what image had been used in the facial recognition search. The department has since prohibited that use of the system, according to ABC News.
That matters. So does what remains unknown.
It is not yet clear how much weight the AI result carried in the arrest decision. It is not yet clear which officials saw the evidence that pointed away from Lipps. It is not yet clear how the safeguards failed at each step.
But a safeguard that only works after months in jail is not much of a safeguard.
This case also shows why “human review” can mean very little if the human treats a machine result as an answer. A person may sign the paper, make the call, and own the decision. That does not help if the system trains people to stop asking basic questions.
Was she there? Did the witness identify her? Did the records place her near the crime? Did the images match? Did the supposed match come from a real suspect, or from someone’s fake ID?
Those are not advanced AI questions. They are ordinary investigation questions.
Lipps was eventually released on Christmas Eve after bank records showed she was in Tennessee when the crimes occurred, according to the lawsuit and news reports. The complaint says she lost her residence, car, and other property during the detention.
The case now sits where many technology scandals end up: in court, after the damage.
That order should bother everyone. A lawsuit can seek money. A policy can ban one use of a tool. Neither can give back the months taken from a person who was caring for four children on an ordinary day.
The hardest fact here is not that AI can be wrong. Everyone already knows that.
It is that a wrong result can look official long enough to put an innocent person on the ground.
Browse: News