TL;DR: Police across the U.S. have arrested innocent people based on facial recognition matches at least 13 times, and those are just the cases we know about. Angela Lipps, a 50-year-old Tennessee grandmother, spent five months in jail for North Dakota bank fraud she couldn't have committed. She was 1,200 miles away. Porcha Woodruff was arrested for carjacking while seven months pregnant. Robert Williams was cuffed in his driveway in front of his daughters. Nearly every confirmed victim is Black. The technology misidentifies darker-skinned faces at rates 10 to 100 times higher than white faces. Police departments know this. They use it anyway. Here's every known case, the racial data behind the failures, and why it's getting worse.
Five Months for a Crime 1,200 Miles Away
On July 14, 2025, U.S. Marshals showed up at Angela Lipps' home in Carter County, Tennessee. She was babysitting. They arrested her at gunpoint [1].
Lipps, a 50-year-old grandmother, learned she'd been charged with four counts of bank fraud and four counts of identity theft in North Dakota. The charges stemmed from someone using a fake U.S. Army military ID to withdraw thousands of dollars from banks in the Fargo area between April and May 2025 [2].
The problem: Lipps had never been to North Dakota. Not once.
The Fargo Police Department doesn't even have its own facial recognition system. Neighboring West Fargo does. Their system flagged Lipps as a "potential suspect" based on a photo from the fake ID. A detective reviewed her social media and decided that was enough for a warrant [2]. He never checked whether Lipps was actually in North Dakota during the fraud. He didn't request her bank records. He didn't verify her location.
Lipps sat in a Tennessee jail for 108 days as a fugitive, held without bail. Officers from North Dakota didn't pick her up until October 30. Once she finally arrived in Fargo and her attorney Eric Rice requested her bank records, the case collapsed in a single meeting on December 19. Her records showed she was at home in Tennessee, 1,200 miles away, when police said she was committing fraud in Fargo [1].
By then, she'd lost five months. According to her attorney, she also lost her home, her car, and her dog [3].
Fargo police haven't apologized [4].
The Pattern Nobody Wants to Talk About
Lipps isn't an outlier. She's the latest entry in a pattern that NBC News documented in March 2026: at least 13 criminal cases have been dismissed nationwide after facial recognition matches identified the wrong person [5]. AI watchdog organizations say the errors are accelerating.
Here's what makes this pattern impossible to ignore: nearly every confirmed victim is Black.
That's not a coincidence. The National Institute of Standards and Technology tested 189 facial recognition algorithms in 2019 and found that Black and Asian faces were misidentified at rates 10 to 100 times higher than white faces [6]. Detroit Police Chief James Craig acknowledged that his department's system produced wrong results 96% of the time when used as the sole identification method [6].
Ninety-six percent. And Detroit kept using it.
Every Known Case
These are the wrongful facial recognition arrests that have been publicly documented. There are almost certainly more: police departments routinely hide their use of the technology, sometimes listing facial recognition results as tips from a "credible source" in arrest warrants [7].
Robert Williams, Detroit, January 2020
The first publicly reported case. Detroit police arrested Williams in his driveway, in front of his two young daughters and his wife, for a shoplifting incident. Officers used a blurry surveillance still to run a facial recognition search, got a match, and showed up with handcuffs. Williams spent 30 hours in an overcrowded cell. When a detective showed him the surveillance photo during interrogation, Williams held it up next to his face and said, "I hope you don't think all Black men look alike." The detective said, "The computer says it's you" [8]. The ACLU sued. The case was settled.
Michael Oliver, Detroit, 2019
Wrongfully arrested for the same reason as Williams: Detroit police ran a facial recognition search, got a match, and treated it as a positive ID. Oliver was charged. The case was eventually dismissed [8].
Nijeer Parks, Woodbridge, New Jersey, 2019
Police were so intent on building a case against Parks after a facial recognition match that they ignored DNA and fingerprint evidence pointing to someone else. Parks was accused of shoplifting and assault. He spent 10 days in jail and nearly three years fighting the charges. Woodbridge settled for $300,000 without admitting wrongdoing [8].
Porcha Woodruff, Detroit, 2023
Arrested for carjacking while seven months pregnant. Nothing in the surveillance images or witness statements indicated the suspect was pregnant. Detroit police ran a facial recognition search, got a match, and sent officers. Woodruff was interrogated for 11 hours at the Detroit Detention Center. The case was dismissed after one month [6].
Randal Quran Reid, Jefferson Parish, Louisiana, 2022
Pulled over outside Atlanta the day after Thanksgiving. Told he was wanted in Louisiana for stealing luxury purses from a consignment shop in Metairie. The arrest warrant listed the identification source as "a credible source," with no mention of facial recognition, no mention of Clearview AI [7]. Reid, a Georgia resident, spent six days in jail before the warrant was rescinded. He'd never been to Louisiana. Jefferson Parish settled for $200,000 [9].
Harvey Eugene Murphy Jr., Houston, 2024
Filed a lawsuit alleging wrongful arrest based on a facial recognition match by Houston police. Case details emerged through legal filings [8].
Jason Vernau, Miami
Spent three days in jail after being accused of cashing a fraudulent $36,000 check at a Miami bank. Police relied on a facial recognition match of the wrong person [8].
Alonzo Sawyer, Details Emerging
Another Black man wrongfully accused based on a facial recognition match. His case is among the confirmed wrongful identifications documented by civil rights organizations [8].
The remaining cases among the 13 documented dismissals haven't been fully reported. That's partly because police departments don't have to disclose when they use facial recognition, and partly because defendants often don't know the technology was used against them until discovery in their cases, if they get that far [8].
Why "It's Just a Lead" Doesn't Work
Every police department that uses facial recognition has a policy that says the same thing: results are investigative leads, not positive identifications. The International Association of Chiefs of Police says a match is "a strong clue, and nothing more, which must then be corroborated against other facts" [10]. Detroit's own policy, adopted in September 2019, states in capital letters that results are "NOT TO BE CONSIDERED A POSITIVE IDENTIFICATION" [10].
The ACLU identified two reasons these warnings are worthless in practice [10]:
Officers ignore them. Louisiana officers arrested Randal Quran Reid based solely on a Clearview AI match. They didn't verify his location. They didn't check whether he'd ever been to Louisiana. They just wrote "credible source" on the warrant.
The follow-up investigation is contaminated. When police use a facial recognition match to assemble a photo lineup, the algorithm's pick often looks more similar to the suspect image than the fillers do. That biases witnesses toward confirming the machine's choice. This is exactly what happened in the Detroit cases: Williams, Oliver, and Woodruff were all selected in lineups that started with a facial recognition match [10].
The result: a policy that exists on paper but fails every time it matters.
The Numbers Behind the Failures
The scale of facial recognition in American policing is staggering, and the safeguards are close to nonexistent:
- Half of U.S. adults have their photos in law enforcement facial recognition databases, according to Georgetown University's 2016 study [6].
- 200+ million images sit in government facial recognition systems used by agencies from local police to ICE [5].
- 10 to 100x higher error rates for Black and Asian faces compared to white faces, per NIST testing of 189 algorithms [6].
- 96% error rate acknowledged by Detroit's own police chief when the system is used alone [6].
- $3.75 million: ICE's largest Clearview AI contract ever, signed in 2025 [11].
- 100,000+ searches run by ICE agents using the Mobile Fortify facial recognition app [11].
The algorithms struggle with darker skin tones because the training datasets are overwhelmingly composed of lighter-skinned faces. The systems can't distinguish between facial structures when the contrast is lower. That's not a bug being fixed. It's a structural flaw in how the technology works [6].
Who's Trying to Stop This
The legislative response is growing, but it's fragmented:
Federal level: Senator Ed Markey, along with Senators Merkley and Wyden and Representative Jayapal, introduced the ICE Out of Our Faces Act in February 2026. It would ban ICE and CBP from using facial recognition entirely and require deletion of all biometric data [12]. It's endorsed by the EFF, ACLU, EPIC, and a dozen other organizations. It hasn't received a vote.
New York: The State Senate passed the Facial Recognition Technology Study Act (S3699) in March 2026, establishing a task force to study how the technology is used. New York City also has separate bills targeting facial recognition by landlords and businesses [13].
City bans: Over 20 jurisdictions (including Boston, San Francisco, and Pittsburgh) have banned police from using facial recognition. These bans work. No wrongful arrest has been reported in a city with an active ban [10].
Innocence Project: Calling for a moratorium on facial recognition in criminal cases, mandatory disclosure to defense attorneys when the technology is used, and rigorous courtroom testing of AI evidence [6].
The gap between where the law is and where the technology is being deployed gets wider every month.
What You Can Do
Find Out If Your City Uses It
File a public records request with your local police department asking whether they use facial recognition tools, which vendor they use, and how many times they've run searches in the past year. Many departments won't tell you voluntarily.
Support Local Bans
City-level bans work. Over 20 cities have passed them. If your city hasn't, contact your city council and point them to the documented wrongful arrest cases. The ACLU has a toolkit for advocacy.
Know Your Rights If Arrested
If you're arrested and suspect facial recognition was involved, your attorney should file for discovery requesting all investigative tools used. Police often hide facial recognition behind vague language like "credible source" or "investigative lead." Ask specifically.
Limit Your Facial Data
Opt out of platforms that feed facial recognition databases. Remove your photos from data broker sites. Avoid uploading clear facial photos to public social media. West Fargo police used Angela Lipps' social media to "confirm" a facial recognition match that was completely wrong [2].
The Machine Keeps Running
Thirteen case dismissals should have been a wake-up call. It hasn't been. Police departments keep buying the technology. Clearview AI keeps signing contracts. Its $3.75 million ICE deal is its largest ever. The federal government is expanding, not restricting, its use of facial recognition.
Angela Lipps lost five months of her life because a computer in West Fargo thought she looked like someone else. Porcha Woodruff was dragged to an interrogation room while seven months pregnant. Robert Williams was handcuffed in front of his children. Randal Quran Reid spent Thanksgiving weekend in jail for crimes in a state he'd never visited.
They were all innocent. They were almost all Black. And the system that failed them is still running, right now, searching faces against databases of 200 million images, producing "leads" that police treat as proof.
Thirteen dismissed cases isn't a failure rate. It's a body count. And it's only the number we know about.
References
- CNN: Police used AI facial recognition to arrest a Tennessee woman for crimes committed in a state she says she's never visited (March 29, 2026)
- NewsChannel 9: Tennessee grandma mistakenly sent to North Dakota jail due to AI error (2026)
- Jezebel: I Hope You're Ready to Spend Months in Jail After AI Facial Recognition Tools Frame You for a Crime (2026)
- Reason: Fargo Police Refuse to Apologize to Tennessee Grandma Jailed on Bogus AI Evidence (March 30, 2026)
- NBC News: 'It's a Wild West': AI watchdogs say facial recognition policing errors are on the rise (March 2026)
- Innocence Project: When Artificial Intelligence Gets It Wrong
- Seattle Times: Police relied on hidden technology and put the wrong person in jail
- Washington Post: Arrested by AI: Police ignore standards after facial recognition matches (2025)
- Biometric Update: Wrongful arrest linked to facial recognition error leads to $200K settlement
- ACLU: Police Say a Simple Warning Will Prevent Face Recognition Wrongful Arrests. That's Just Not True.
- NBC News: ICE used facial recognition app more than 100,000 times (2026)
- Senator Markey: ICE Out of Our Faces Act (February 5, 2026)
- NY Senate: Passage of Facial Recognition Technology Study Act (2026)
Published: April 11, 2026