Falsely Accused by AI and Jailed for 160 Days: American Woman Steps into the Execution Zone

marsbitPublished on 2026-04-27Last updated on 2026-04-27

Abstract

Summary: Angela Lipps, a woman from Tennessee, was wrongfully imprisoned for 160 days after a facial recognition AI incorrectly identified her as a suspect in a bank fraud case in North Dakota. Despite her insistence that she had never been to the state, she was arrested in July 2025 and held without bail. She spent 108 days in a local jail before being transferred to North Dakota in October, where she was eventually cleared after her lawyer provided bank records proving she was buying pizza in Tennessee at the time of the crime. Her ordeal highlights a growing issue with AI-driven law enforcement errors. Similar cases include Robert Williams, detained for 30 hours in 2020, and Chris Gatlin, jailed for 17 months based on flawed AI matches. Systems like Clearview AI, used by over 3,000 U.S. law enforcement agencies, often rely on low-quality images and can produce inaccurate results. Human oversight failures exacerbate the problem, as officers sometimes skip verification steps, leading to wrongful arrests. While AI has improved policing efficiency, its errors have serious consequences, emphasizing the need for stricter regulations and human accountability in its use.

What would you do if AI suddenly declared you guilty, and everyone blindly believed its judgment? How would you prove your innocence?

This sounds like a plot from the movie "Minority Report," where crimes are predicted and people are framed. But abstractly, a sloppier version has already played out in reality:

AI misidentified a criminal, leading an innocent woman to be jailed for half a year, nearly wiping out her life.

1. What does it mean for the police to arrest "The Flash"?

Angela Lipps from Tennessee, USA, is this unlucky woman. In July 2025, a team of heavily armed police suddenly burst into her平淡的生活 (ordinary life) and pointed guns at her, informing her she was under arrest.

She was stunned, as she had no idea what she had done wrong. The police said she was suspected of involvement in a bank fraud case in North Dakota. It seemed like they had solid evidence and had issued a 5-star warrant. But after hearing the reason, she froze in place:

"But, I've never been to that state in my life."

Looking at the map, there is indeed some distance between Tennessee and North Dakota. What connected her to the suspect was AI.

The police ignored her protests when arresting her. She only learned the specific case details after being in prison. The police in Fargo, North Dakota, were investigating a fraud case where the suspect used a fake military identity to scam tens of thousands of dollars from a bank. The police's investigation method was to first check surveillance footage, then consult AI, and finally make an arrest.

You might think they used AI to refine a Holmes-like detective who solves cases through rigorous logical reasoning, but it wasn't that "complicated." The police used an AI system to perform facial recognition on screenshots from the surveillance video and then filtered for suspects in the database.

This time, the AI dropped the ball big time, incorrectly matching Angela Lipps's photo. Then this intelligence was transferred to the investigating police station, where the staff made an even bigger mess. After manually checking her driver's license and social media, they thought the facial features, body shape, and hairstyle were similar, and thus determined she was the suspect.

Image from GoFundMe

The woman was utterly崩溃 (crushed). Didn't they check other clues? What if it was just a case of lookalikes? But what was even more崩溃 (crushing) was what came next: after a series of moves, she had no chance to prove her innocence.

According to local procedures, she wasn't directly taken to North Dakota for further investigation but was instead held in a local Tennessee jail. Because she was identified as an "interstate fugitive," she had no bail and no opportunity for police interrogation. She was just inexplicably detained for 108 days, from July until the end of October. It's hard to imagine how desperate she felt during this time.

Angela Lipps endured until October 30th, when she was finally taken out of jail to be transferred to North Dakota. At that time, she faced multiple charges. During the trip, she not only had to wear handcuffs but also had a chain around her waist. To her, walking through the airport was like being paraded through the streets.

And this was actually her first time ever on a plane.

What she didn't expect was that after arriving in North Dakota, she had to wait even more. It wasn't until December that she had a formal chance to explain the matter. Her lawyer directly subpoenaed her bank records to compare the timing. At the time of the crime, she was buying pizza 1900 kilometers away in Tennessee. This meant that unless she was "The Flash," she couldn't have committed the crime across spacetime.

It took only 5 minutes for the lawyer to clear things up for her.

By December 24th, her charges were finally dropped, but she couldn't be happy, because by then the "execution line" was already at her ankles.

Angela Lipps spent her 50th birthday in jail. During her imprisonment, she wasn't allowed to wear her dentures, ate junk food, and was under immense mental stress, causing her weight to skyrocket. She entered jail in summer clothes and came out in winter, without even a warm coat, making staying warm a problem, let alone getting home. The police didn't provide any travel money either. After being released, she was essentially stranded in North Dakota, trapped on the day AI wrongfully accused her.

Fortunately, some kind souls gave her a helping hand.

Her lawyer helped arrange a hotel for temporary stay and got her some food to get through the toughest moments. Local non-profit organizations came to her rescue and sent her back home to Tennessee.

But after returning home, restarting her life was hellishly difficult. During her time in jail, her finances collapsed: her rented apartment was gone, her dog was gone, and even her car and the things inside it were gone. Her personal belongings were temporarily stored in a warehouse, but later, because she couldn't afford the rent, all her daily necessities were cleared out. Even her interpersonal relationships were in jeopardy because the neighbors saw her being taken away and "disappearing" for half a year, thinking she had actually committed a crime.

Meanwhile, her bizarre experience quickly spread online, sparking疯狂吐槽 (furious吐槽/roasting). For example, many expressed disbelief that in the US, some people can commit "zero-dollar shopping" (looting without consequence), but if you touch a capitalist's bank, you can be arrested instantly, even if you didn't do it.

Because the incident blew up, her desperate situation took another turn for the better, and she received more help. She has now crowdfunded $80,000 in donations, which can fully be used to restart her life.

However, netizens' biggest expectation now is for her to find a dream team of lawyers and sue them to high heaven, potentially winning the case and receiving another compensation payout.

Because this wasn't an偶然失误 (accidental mistake) by the police; batch wrongful convictions have occurred before this, and some people's experiences were even worse than hers.

class="image-wrapper">

2. The Chilling Sci-Fi Moments Brought by AI "Mistakes"

Angela Lipps's experience has actually happened over a dozen times in different regions, even in other countries.

For example, Robert Williams was one of the earliest victims in this type of incident. In 2020, police arrested him in front of his wife and daughter. He was detained for 30 hours, even though he was 20 centimeters taller than the actual perpetrator.

Another top倒霉蛋 (unlucky person) is Chris Gatlin. AI matched him, based on blurry surveillance footage, to a potential suspect in a subway assault case. He was subsequently locked up for 17 months straight, becoming the longest-detained innocent person so far. The离谱 (absurd) part is that until the very end, the investigators discovered there was also a bodycam recording that could serve as key evidence, but the suspect in it didn't match him at all.

Porcha Woodruff in 2023 had a similar experience. Police accused her of being involved in a carjacking. She laughed out loud upon hearing it, thinking the police were joking, because she was eight months pregnant at the time, and clearly didn't seem to be in any condition to commit such an act. Nevertheless, she was taken away and held for over ten hours. Later, she lost her lawsuit.

This year, a young man named Alvi Choudhury in the UK was also arrested under similar circumstances, accused of "burglary," but the evidence suggested it was done "remotely"—a scene also documented in film. The AI identified his photo from a 2021 detention. The police laughed out loud when they saw him in person because he looked at least 10 years older than the suspect.

Image from "Jumper"

In 2022, Indian businessman Praveen Kumar was arrested on his way to Switzerland. While transiting through Abu Dhabi, AI matched him to a wanted criminal. The local authorities interrogated him thoroughly, finally realizing he indeed wasn't the suspect, and deported him back to India. However, upon arrival at the Indian airport, he was detained twice more consecutively.

Among these incidents, the most离谱 (absurd) might be the experience of Russian scientist Alexander Tsvetkov. AI said he looked like a murderer, and he was jailed for 10 months.

This scientist suffered from双重诬陷 (double framing) by both AI and humans. In February 2023, he returned to Russia after a scientific expedition, only to be arrested right after getting off the plane. The reason was that AI showed a 55% similarity between him and a sketch of a suspect from a serial murder case over twenty years ago. Additionally, there was a tainted witness in the case who deliberately gave false testimony against him to reduce his own sentence. The致命 (fatal) part was that the police didn't investigate carefully and just arrested him.

After being detained, he initially thought it was a small misunderstanding, but the investigation dragged on endlessly. Under immense mental pressure and health issues, he was once compelled to plead guilty, only to recant later. Fortunately, his wife and colleagues from the research institute persistently worked for him, finding countless materials proving he was on an expedition elsewhere at the time of the crime. As media reported on the case, it sparked widespread public discussion. The situation only took a turn in December, and the charges weren't formally dropped until February 2024.

And as AI evolves, this type of case is happening more and more frequently, making it seem more like AI is "commanding," while some humans have become execution machines.

3. The Occupational Hazard of "Letting AI Decide"

The reason it feels abstractly like a "role reversal" is that both AI and humans can drop the ball. Even the strongest AI makes mistakes, and humans are always ready to "slack off."

Take the American Clearview AI system, for example—it's the one that sent the woman from the beginning of the article to jail, but it's no amateur software. Clearview AI boasts the world's largest image database, aggressively scraping tens of billions of images despite fines from multiple countries. Over 3,000 law enforcement agencies in the US use it, and last year it even signed a $9.2 million contract with ICE.

Simply put, Clearview AI is a "facial search engine." You upload a picture, it matches it within its database, and then spits out a bunch of similar images. Ultimately, whether it's the person you're looking for or not is still up to you to decide.

Although AI has made神速 (rapid) progress these years, its "accuracy rate" remains highly controversial, as there are always various differences between experimental test data and real-world usage when searching with images.

For instance, most of the time, the surveillance images the police find have the picture quality of a landline phone shot (extremely low resolution). Add factors like weird lighting, tricky angles, facial obstructions, etc., making recognition quite difficult. The database might also contain old pictures, and the matching results might just be two people who happen to look alike.

And when these errors are applied to specific individuals, they naturally suffer.

Is this a crow or a cat?

In theory, if it were just AI occasionally malfunctioning, there might be a way to fix it, as these are just "clues." The subsequent "conclusion" still needs human verification. But often, humans are more "离谱 (absurd)" than AI.

Actually, people have been complaining for years that American police are being led by the nose by AI, falling into "automation bias psychology"—meaning they rely on or even over-trust the results given by AI. If AI says these two people match, humans believe it. Height doesn't match? Not my problem. Alibi exists? I just make arrests. What other evidence? Tell it to the prison. They skip basic investigation steps while making others "skip life."

Humans are actually well aware of AI's shortcomings. After paying quite a bit in compensation, many regions in the US have established "firewalls," such as requiring independent evidence besides AI clues for solving cases. Some places have even outright banned the use of this type of AI technology in investigations.

But these constraints can't withstand how incredibly useful AI is. Although there's a probability of failure, clearly the number of successes is higher. Sometimes, the only clue the police painstakingly find is a surveillance photo of "door lock" quality (very poor), so there's a high probability they just want to try their luck with AI. As early as 2023, there were reports that US police had tried over a million searches using Clearview AI. Even later, when explicitly told not to use it, they used it secretly anyway. After all, they could always claim they didn't use it. If this software wasn't allowed, they'd switch to another. If it was banned locally, they'd ask another agency to help use it. They're utterly addicted to using it.

So, once AI first provides an unreliable clue, and then humans slack off by skipping the investigation, the dynamic duo unleashes a combo move, and wrongful convictions naturally follow.

For example, Brazil's Smart Sampa system has been facing similar issues in recent years. In 2024, São Paulo, Brazil, launched Latin America's largest AI facial recognition policing system, reportedly connected to 40,000 cameras.

The good news is that the effect is indeed fierce. Over the years, nearly four thousand criminals were arrested on the spot, and over three thousand fugitives were captured. Robberies dropped by nearly 15% in 2025,堪称 (can be called) an "assembly line for catching thieves."

The bad news is that at least 59 people were misidentified.

Among them were abstract cases, such as a mentally ill person being taken from a hospital as a criminal. Later, it was discovered his arrest warrant had expired, and he was released. Another guy was arrested 4 times in 7 months because AI confused him with a fugitive murderer. Each time, he was taken to the police station and released immediately, only to be arrested again a few days later. He was utterly terrified.

When we used to discuss which jobs couldn't be replaced by AI, we said AI couldn't serve prison time for people. But we never imagined the other side of this question: it can now make people "serve prison time."

This statement was a joke in the past, but now it's almost not news anymore. Actually, it's not just facial recognition; AI can also pull off some狠活儿 (tough stunts) in object recognition occasionally.

Last year, an AI security system at a US high school识别 (recognized) a bag of chips a young man was holding as a "potential gun," directly triggering an alarm. Eight police cars arrived and instantly restrained the young man. After searching him, they found the snack bag in the trash. The scene was extremely awkward, and the young man thought he was done for.

You have to admit, it kinda does

Who would have thought that the stronger AI gets, the bigger the mess some people can make with it. In the past, AI was dumb; it misidentified pictures and was laughed at by the whole internet. Now AI is strong; misidentifying one person lands a human in jail for half a year. In the future, let's not end up with abstract operations like AI judgment, AI lawyers, or新能源法庭 (new energy courts). Ultimately, AI itself is just a tool. The key lies in the ability and purpose of its users. Or rather, the stronger AI becomes, the less humans can afford to slack off.

In short, I'm starting to miss that dumb AI a little.

This article is from the WeChat public account "Cool Play Lab," author: Cool Play Lab

Related Questions

QWhat was the main reason Angela Lipps was wrongfully arrested and imprisoned for 160 days?

AShe was wrongfully arrested and imprisoned because an AI facial recognition system (Clearview AI) incorrectly matched her to a suspect in a bank fraud case in North Dakota, despite her never having visited that state.

QHow did Angela Lipps finally prove her innocence in the bank fraud case?

AHer lawyer used her bank records to show that at the time of the crime, she was buying pizza in Tennessee, 1900 miles away, making it impossible for her to have committed the crime in North Dakota.

QWhat are some of the personal consequences Angela Lipps faced after her wrongful imprisonment?

AShe lost her rented home, her dog, her car and personal belongings stored in a warehouse, suffered weight gain and mental stress, and faced damaged relationships with neighbors who thought she was guilty.

QName one other individual mentioned in the article who was wrongfully arrested due to AI facial recognition errors.

ARobert Williams was an early victim, wrongfully arrested in 2020 and detained for 30 hours after an AI mismatch, despite being 20 centimeters taller than the actual suspect.

QWhat is the 'automation bias' mentioned in the article in the context of law enforcement using AI?

AAutomation bias refers to the tendency of law enforcement to over-rely on or overly trust AI-generated results, sometimes skipping essential investigation steps and leading to wrongful arrests based solely on AI matches.

Related Reads

Deconstructing the U.S. Stock Quantum Computing Sector: IonQ, Rigetti, D-Wave, Which of These Concept Stocks is Worth Betting On?

**Title:** Analyzing the US Quantum Computing Race: IonQ, Rigetti, D-Wave – Which Concept Stock is Worth Betting On? **Summary:** The podcast discusses the resurgence of quantum computing as a national priority for both the US and China, driven by its potential to break current encryption, revolutionize drug discovery, finance, and logistics. The core challenge is commercializing the technology, which is hampered by high error rates in quantum bits (qubits). Quantum error correction, requiring thousands of physical qubits per reliable logical qubit, is key but years away. The analysis compares three main publicly traded US quantum computing firms: * **IonQ (Ion Trap):** Considered the most financially stable with the fastest commercial progress (2025 revenue: $130M, +202%) and high-quality clients. Its valuation is very high, pricing in significant future growth. * **Rigetti (Superconducting):** Seen as the highest-risk, highest-potential-reward bet. It has the smallest revenue but recently launched a 108-qubit system. Its valuation multiples are extreme, making it highly sensitive to news. * **D-Wave (Quantum Annealing):** Has the most unique positioning with real-world enterprise clients today (e.g., Mastercard, Volkswagen) solving optimization problems. Its recent acquisition moves it into general-purpose quantum computing ("dual-platform"), adding execution risk. Major tech giants like Google, IBM, and Microsoft are also heavily invested, pursuing various technical approaches. Nvidia is positioning itself as the essential bridge between classical and quantum computing. The investment phase is likened to AI in 2018-2020: promising underlying technology with accelerating breakthroughs but a commercial inflection point still 3-7 years away, suggesting potential for a market correction ("bubble washout"). For investors, suggested approaches include gaining exposure through tech giants with quantum divisions (e.g., Google, IBM) or using niche ETFs like WQTM for pure-play quantum exposure, rather than direct stock picks in the highly volatile pure-play companies at this early stage.

marsbit12m ago

Deconstructing the U.S. Stock Quantum Computing Sector: IonQ, Rigetti, D-Wave, Which of These Concept Stocks is Worth Betting On?

marsbit12m ago

From Parallel Finance to Mainstream Finance: The On-Chain Securities Era Ushers in a Historic Window

From Parallel Finance to Mainstream: The Dawn of On-Chain Securities For over a decade, the crypto industry has operated as a parallel financial system with its own currencies, markets, and assets—from Bitcoin and ICOs to DeFi, NFTs, and memecoins. Despite building a robust internal ecosystem, a wall has separated it from the traditional financial world. That barrier is now crumbling. The industry's first act was one of internal evolution: ICOs streamlined fundraising, DeFi recreated financial services on-chain, and layer-2 networks competed for scalability—all within the crypto bubble. While innovative, this cycle remained closed, with capital and users circulating internally, leading to volatile boom-bust cycles. Even Bitcoin ETFs, while attracting Wall Street capital, merely provided a channel to buy crypto assets without bridging the systems. The next, larger narrative is Real-World Assets (RWA) moving on-chain. This involves tokenizing stocks, bonds, funds, and future cash flows. Blockchain can compress the complex traditional processes of trading, settlement, clearing, and custody into a seamless, automated network operating in seconds. This shift is creating a new financial gateway: the native crypto securities broker. This entity will combine functions of an exchange, broker, bank, and custodian into a unified global financial operating system. Consequently, the next major battleground won't be the "public chain wars" focused on speed and cost, but the competition to build the financial infrastructure capable of hosting high-quality, liquid real-world assets. Access to global equities, index funds, or stakes in companies like SpaceX could erase the boundary between crypto and traditional finance, unlocking a market orders of magnitude larger than crypto's current valuation. In summary, after years of creating a separate financial world, crypto's next decade will be defined by its integration into the existing global financial system, marking the true beginning of its largest growth story.

marsbit33m ago

From Parallel Finance to Mainstream Finance: The On-Chain Securities Era Ushers in a Historic Window

marsbit33m ago

Wang Chuan: When the Neighbor Old Wang Made 30x on Memory Stocks, How to Avoid Anxiety (Part Six) - The Trap of Commoditized Goods

Wang Chuan: When the Neighbor Lao Wang Made 30x on Storage Stocks, How to Stay Anxiety-Free (Part 6) - The Trap of Commoditized Goods. This essay uses historical and current examples to analyze the cyclical and high-risk nature of the data storage industry. It begins with the 1990s rise and dramatic fall of Iomega, whose stock soared over 160x in 18 months before collapsing 97% from its peak, illustrating the fleeting success of storage "meme stocks." The core problem is that storage products, like DRAM and flash memory, are highly commoditized. This leads to extreme volatility: prices have plummeted over 80% multiple times, and company stocks often crash 95% or go bankrupt. The industry's dynamic is defined by "elastic demand facing heavy-asset, long-cycle, rigid supply." When demand spikes and supply is fixed, prices skyrocket, as seen recently with AI-driven demand for High Bandwidth Memory (HBM). Companies like Sandisk and Micron have reported massive revenue and gross margin jumps (e.g., Sandisk's gross margin rising from 22.5% to 78.3%) despite minimal increases in production volume. However, these high margins are self-defeating. They incentivize massive new capacity investments (hundreds of billions planned from 2026), with supply expected to surge by late 2027. Once new supply meets demand, prices and profits will crash, potentially leading to a scenario where "selling more results in earning less." The article debunks the safety of long-term supply agreements, comparing them to fragile non-aggression pacts easily broken when market conditions shift. It warns that when an industry is highly profitable but trades at low P/E ratios, the risk is greatest, as plummeting prices quickly erase those earnings. Multiple asymmetric risks loom, including economic recession, reduced AI spending, faster-than-expected capacity expansion (especially from Chinese firms), and technological innovations that reduce memory requirements. In conclusion, the storage sector is a cyclical trap where periods of euphoric profits are often precursors to devastating downturns, luring unprepared investors into a "wealth incinerator."

marsbit42m ago

Wang Chuan: When the Neighbor Old Wang Made 30x on Memory Stocks, How to Avoid Anxiety (Part Six) - The Trap of Commoditized Goods

marsbit42m ago

Wang Chuan: When the neighbor Lao Wang earned thirty times from investing in memory storage stocks, how can you still avoid anxiety (6) - The trap of homogeneous products

The article, "Wang Chuan: How to Remain Unanxious After Neighbor Lao Wang's Thirty-Fold Gain on Storage Stocks (Part 6) - The Trap of Commoditized Goods," analyzes the cyclical and perilous nature of the data storage industry through historical and current case studies. It begins with the example of Iomega, whose Zip drives led to a stock surge of over 160x in the mid-1990s before collapsing over 97% from its peak due to competition from cheaper CD-R technology. This pattern is characteristic of storage, where products like DRAM are highly commoditized, leading to extreme price volatility. The sector has seen prices crash over 80% multiple times, with companies often facing bankruptcy. The core dynamic is "elastic demand facing heavy-asset, long-cycle, rigid supply." High prices attract new capacity, but the long lead time means supply eventually overshoots, causing sharp price corrections. The current AI-driven boom, exemplified by surging demand for High-Bandwidth Memory (HBM), has led to skyrocketing prices and profit margins for companies like SanDisk and Micron, despite relatively flat production volumes. However, the author warns this high-margin environment is self-defeating. The high profits are already triggering massive new capacity investments (hundreds of billions starting 2026), with supply expected to ramp up by late 2027. When supply catches up, total revenue and profits may fall even as more units are sold. Long-term supply agreements offer little protection, as buyers can find ways to renegotiate if market prices drop, similar to fragile political treaties. Key risks include economic downturns, cuts in AI spending, faster-than-expected capacity expansion (especially from Chinese firms), and innovations in chip/algorithm design that reduce memory needs. A critical trap is that at the cycle's peak, storage stocks often appear cheap with low P/E ratios, luring value investors just before an impending downturn where profits evaporate. The conclusion cautions that for commoditized goods like storage, high margins inevitably destroy themselves, and the current asymmetry favors downside risk over further upside. The neighbor's dream of easy wealth from storage stocks is portrayed as a precarious illusion.

链捕手1h ago

Wang Chuan: When the neighbor Lao Wang earned thirty times from investing in memory storage stocks, how can you still avoid anxiety (6) - The trap of homogeneous products

链捕手1h ago

Trading

Spot
Futures

Hot Articles

Discussions

Welcome to the HTX Community. Here, you can stay informed about the latest platform developments and gain access to professional market insights. Users' opinions on the price of AI (AI) are presented below.

活动图片