TL;DR: In April 2025, the NYPD arrested Trevis Williams, a 36-year-old Brooklyn man, for a sex crime in Union Square, Manhattan. The basis: a facial recognition match. The problem: Williams was driving from Connecticut to Brooklyn when the crime occurred: his cell phone records prove it. The real suspect was eight inches shorter and 70 pounds lighter. The only similarity? Both men are Black with locs. Prosecutors dropped the case after the Legal Aid Society proved the identification was false. Williams had been in the middle of a hiring process to work as a correctional officer at Rikers Island. That process is now frozen. He’s the 14th person known to have been wrongfully arrested in the U.S. because police trusted a facial recognition match over basic investigative work.
A Sex Crime He Couldn’t Have Committed
The crime happened in Union Square, Manhattan. Someone exposed themselves to a woman. The NYPD ran surveillance footage through its facial recognition system. The algorithm returned a match: Trevis Williams [1].
Williams was arrested and held for two days. Two days in a cell for a sex crime he didn’t commit, in a place he wasn’t at, committed by a person who looked nothing like him [1].
His cell phone location data showed he was driving from Connecticut to Brooklyn at the time. The actual suspect, who was later photographed committing the crime, was eight inches shorter and 70 pounds lighter than Williams. Both men were Black. Both had locs. That was the extent of the resemblance [1] [2].
The NYPD treated the algorithm’s output as probable cause. Nobody checked height. Nobody checked weight. Nobody checked where Williams actually was.
A Career Frozen by an Algorithm
Williams wasn’t just any Brooklyn resident who got swept up. He was in the process of being hired as a correctional officer at Rikers Island. The arrest froze that hiring process [1].
Think about that. A man applying to work inside the criminal justice system had his career derailed by the criminal justice system’s own technology. The same facial recognition tools his future employer uses to “keep people safe” destroyed his job prospects based on a match so wrong that the suspect didn’t even share his height or build.
Prosecutors dismissed the case after Williams’s public defenders at the Legal Aid Society proved the identification was false [1]. The Legal Aid Society then sent letters to authorities including Inspector General Jeanene Barrett, detailing patterns of false arrests and improper use of facial recognition technology by the NYPD [1].
“The NYPD Cannot and Will Never Make an Arrest Solely Using Facial Recognition”
That’s the department’s official line [1]. Read it again.
They also said facial recognition technology has “proven successful.” For whom? Not for Trevis Williams, who spent two days in jail. Not for his career. Not for the actual suspect, who remained free while police locked up the wrong man.
The NYPD’s claim that it never arrests people “solely” based on facial recognition is a word game. If the facial recognition match is the only lead that generates the arrest, then a photo lineup or brief canvas afterward isn’t independent corroboration: it’s rubber-stamping the algorithm’s output. When the only thing connecting Williams to the crime was a software match, everything that followed was built on that foundation [2].
14 Wrongful Arrests. Same Pattern Every Time.
Williams is the 14th person publicly known to have been wrongfully arrested in the U.S. due to police reliance on faulty facial recognition [3]. The ACLU has documented every case. The pattern is consistent:
- Nijeer Parks: Woodbridge, NJ, February 2019. Arrested for shoplifting and assault. Jailed for 10 days
- Robert Williams: Detroit, MI, January 2020. Arrested on his front lawn in front of his daughters
- Porcha Woodruff: Detroit, MI, February 2023. Arrested while eight months pregnant for a carjacking
- Angela Lipps: Fargo, ND warrant served in Tennessee, July 2025. Armed federal agents arrested her while she was babysitting
- Kimberlee Williams: Maryland, June 2021. Spent six months in jail for bank fraud in a state she says she never visited
Every case was eventually dismissed. Every victim lost something: days, weeks, or months in jail. Jobs. Housing. Their sense of safety. Most of the victims have been Black [3].
Lauren Yu of the ACLU’s Speech, Privacy, and Technology Project put it simply: “No one should spend six months in jail because an algorithm got it wrong” [3].
‘It’s a Wild West’
That’s how AI watchdog organizations described the state of facial recognition in policing to NBC News in April 2026 [4]. The phrase fits.
There are no federal rules governing how police use facial recognition. No national standards for when a match is sufficient for an arrest. No requirement that officers verify the algorithm’s output against basic physical descriptors like height and weight. No mandate to disclose to courts that facial recognition was used [3].
Some cities have banned police facial recognition entirely: more than 20 jurisdictions, including San Francisco, Boston, and Minneapolis. Detroit settled a wrongful arrest lawsuit in 2024 and now prohibits officers from requesting arrest warrants based only on a photo lineup combined with a facial recognition lead [3].
But in New York? The NYPD runs one of the most aggressive facial recognition programs in the country. And when it fails (as it did with Trevis Williams) the department says the technology “has proven successful” and moves on.
The Technology Doesn’t Even Do What They Claim
Facial recognition vendors say their systems are investigative leads, not identifications. Every vendor’s terms of service says the same thing: don’t use this as the sole basis for an arrest.
Police do it anyway.
The ACLU found that in multiple cases, detectives concealed their reliance on facial recognition when obtaining warrants, telling judges they identified a suspect through “investigative means” without mentioning the algorithm [3]. In the Kimberlee Williams case, the detective who sought the arrest warrant never disclosed that the lead came from facial recognition [3].
David Rocah, senior staff attorney at the ACLU of Maryland, called out the inevitable result: “The investigative failures that led to Ms. Williams’ improper arrest are a predictable result” of police reliance on the technology [3].
The technology produces higher false match rates for people of color, women, younger people, and the elderly [3]. It’s least accurate for the populations most likely to encounter police. That’s not a bug. It’s the dataset.
What You Can Do
- Know your rights: If you’re arrested, ask whether facial recognition was used in the investigation. Demand your attorney verify this: police have concealed it before
- Support local bans: Over 20 cities have banned police facial recognition. Check if your city has, and if not, push your city council. The Ban Facial Recognition campaign tracks efforts nationwide
- Support the ACLU: They’ve documented every known wrongful arrest and are pushing for federal legislation requiring independent corroboration before any facial recognition-based arrest
- Contact your congressional representatives: There is still no federal law governing police use of facial recognition. Bills have been introduced. None have passed
References
- ABC7 New York: Man’s wrongful arrest puts NYPD’s use of facial recognition surveillance tech under scrutiny (2026)
- Brill Legal Group: How a Facial Recognition Error Led to a Wrongful Sex Crime Arrest in New York (April 2026)
- ACLU: More than a Dozen Wrongful Arrests Due to Police Reliance on Facial Recognition Technology (2026)
- NBC News: ‘It’s a Wild West’: AI watchdogs say facial recognition policing errors are on the rise (April 2026)
Published: April 22, 2026