Facial Recognition, Bad Detective Work and 163 Days in Jail: The Angela Lipps Case

Imagine being at home in Tennessee, babysitting your neighbor’s children, when U.S. Marshals arrive and arrest you at gunpoint. They tell you that you’re wanted for a series of bank thefts and identity-related crimes committed roughly 1,200 miles away in North Dakota.

You tell them you’ve never been to North Dakota. It doesn’t matter.

You are taken to jail, eventually transported across the country and charged with eight felonies. From the day of your arrest until the charges are dismissed, you spend approximately 163 days in custody.

That is what happened to Angela Lipps, a 50-year-old Tennessee grandmother whose case should concern anyone paying attention to the rapidly expanding use of facial-recognition technology.

Lipps is now suing the City of Fargo and former Fargo police detective Lucas Heck for $10 million, alleging that the improper use of facial-recognition technology, combined with an inadequate investigation, led to her wrongful arrest and prosecution.

The more I read about this case, the more disturbing it became. But this isn't simply another story about artificial intelligence getting something wrong. It is a story about what can happen when humans fail to properly question and verify what technology gives them.

Facial Recognition Produced a Lead, Not Proof

The case originated with an investigation into a series of bank thefts in the Fargo, North Dakota, area. According to Lipps’ federal lawsuit, an investigator with the West Fargo Police Department used facial-recognition software while attempting to identify a suspect.

The technology returned a social media photograph of Lipps as a potential lead. That distinction is extremely important because facial-recognition technology isn't supposed to establish that someone committed a crime. It can provide investigators with a potential match that then needs to be independently investigated and corroborated.

According to the complaint, the image involved in the facial-recognition process came from a fraudulent identification document associated with the woman suspected of committing the crimes. Separately, surveillance cameras captured images of the woman who was actually inside the banks.

The lawsuit alleges there were noticeable differences between Lipps and the woman captured on surveillance, including differences involving their facial features, builds and tattoos. Yet investigators continued pursuing Lipps.

Fargo's former police chief, Dave Zibolski, later publicly acknowledged investigative errors in the case. He explained that Fargo investigators mistakenly believed information received from West Fargo meant that facial recognition had also been performed against the bank surveillance photographs.

It had not.

That may be one of the most important details in this entire story. Identifying the photograph appearing on a fraudulent ID is not the same thing as identifying the human being physically standing inside a bank committing a crime.

From Babysitting in Tennessee to Eight Felony Charges

On July 14, 2025, U.S. Marshals arrested Lipps in Tennessee while she was babysitting four children. She was taken into custody at gunpoint.

Lipps had lived most of her life in Tennessee and maintained that she had never been to North Dakota. According to reporting by The Guardian, she had never even traveled by airplane before authorities eventually flew her there.

She remained incarcerated in Tennessee for nearly four months before being transported to North Dakota, although reporting indicates that unrelated pre-existing probation matters also had to be resolved during that period. Once in North Dakota, Lipps faced four counts of unauthorized use of personal identifying information and four counts of theft. Her bond was reportedly set at $100,000.

From her July arrest until her release on Christmas Eve, Lipps spent approximately 163 days in custody. She spent her 50th birthday behind and Thanksgiving behind bars. She came within hours of spending Christmas there as well.

Basic Evidence Eventually Helped Unravel the Case

This is the part that I find particularly difficult to understand.

According to the lawsuit and subsequent reporting, investigators did not uncover financial records, witnesses or other evidence placing Lipps inside the North Dakota banks. She maintained that she had never been to North Dakota, and the woman shown in surveillance footage allegedly had physical differences that could have raised questions about the facial-recognition lead.

More importantly, Lipps’ attorney eventually obtained bank records showing that she was in Tennessee when the Fargo crimes occurred. That evidence helped dismantle the case against her, and prosecutors ultimately dismissed the charges.

We aren't talking about some futuristic forensic technology that became available months after her arrest. We are talking about ordinary records that could help establish where a person was when a crime occurred.

That is what makes this case about much more than facial recognition. The technology may have generated the original lead, but technology did not carry Lipps through the entire criminal justice process by itself.

Humans did.

At multiple points, investigators had opportunities to ask basic questions: Does she actually resemble the woman captured on surveillance? Has she ever been to North Dakota? Is there financial evidence connecting her to these transactions? Where was she when the crimes occurred? And perhaps most importantly, did the facial-recognition result actually come from the surveillance footage, or was it connected to a photograph appearing on a fraudulent ID?

Those questions don't require artificial intelligence. They require detective work. And this can be applied to any sector of work that uses facial recognition for many purposes.

Released on Christmas Eve With Four Cents

The charges against Lipps were dismissed, and she was released on December 24, 2025. But even her release is difficult to read about.

According to allegations in her lawsuit, Lipps was released into freezing North Dakota weather without a winter coat and without a way to get home to Tennessee. She allegedly had just four cents.

The complaint also describes difficult conditions during her incarceration, including allegations that Lipps went periods without medication and her dentures. By the time she was released, the lawsuit says she had lost her residence, vehicle and personal property.

Those details are allegations contained in ongoing civil litigation and have not all been adjudicated. The City of Fargo has said it does not comment on pending litigation.

Lipps is now seeking $10 million in damages from Fargo and former detective Lucas Heck. Her lawsuit alleges unlawful arrest and imprisonment, malicious prosecution and failures involving the department's policies, training and supervision surrounding facial-recognition technology.

Fargo Has Already Acknowledged Investigative Errors

One reason this case deserves attention beyond the allegations in the new lawsuit is that Fargo's own police leadership previously acknowledged mistakes in the investigation.

Months before Lipps’ lawsuit generated national headlines, Zibolski publicly apologized for investigative errors and announced changes to how the Fargo Police Department would handle facial-recognition information. Among those changes, the department said it would no longer rely on facial-recognition results generated by outside law-enforcement agencies and would implement additional training and safeguards.

Those changes are important. But they came after Angela Lipps had already spent more than five months in custody.

The Problem With Simply Putting a “Human in the Loop”

I am a strong believer in artificial intelligence. Funder Intel uses AI, and I believe the technology will continue making financial services, lending, fraud detection, underwriting, research and countless other industries more efficient.

But supporting AI does not require ignoring its risks. In fact, the more powerful and widespread these systems become, the more seriously we need to examine what happens when they are wrong.

One of the most common assurances surrounding artificial intelligence is that important decisions will remain safe because there will always be a “human in the loop.”

The Angela Lipps case shows why that isn't enough.

There were humans in this loop. Technology produced information, but human beings had to interpret that information, investigate it, decide whether the evidence supported an arrest and determine whether the person identified by the technology was actually the person who committed the crimes.

Having a human involved provides little protection if that human treats a technological output as inherently trustworthy instead of independently verifying it.

This Goes Far Beyond Law Enforcement

The lesson here extends beyond policing. It applies directly to banking, lending, fraud detection, underwriting and compliance. These are areas where increasingly sophisticated AI systems are being asked to identify patterns and make recommendations that can materially affect people's lives and businesses.

Imagine an AI-powered fraud system incorrectly identifying a legitimate business owner as a fraudster. Imagine an underwriting model associating a borrower with someone else's identity or activity. Imagine an automated compliance system producing a false positive and the human reviewer simply clicking “confirm” because the computer must know something they don't.

The consequences aren't always going to involve someone spending 163 days in custody, but the underlying problem is similar.

AI should help investigators investigate. It should help underwriters underwrite. It should help compliance professionals identify risk. It should help human beings make better-informed decisions.

It should never become an excuse to stop thinking.

When an algorithm produces a mistake and humans fail to question it, the consequences stop being technological and become very real for the person on the other side of that decision.

Angela Lipps experienced those consequences for more than five months.

Whatever ultimately happens with her $10 million lawsuit, this case deserves to become part of the larger conversation about how artificial intelligence is used when someone's freedom, livelihood, credit or financial future is on the line.

The question isn't whether AI can make mistakes. We already know it can.

The more important question is what humans do next.