A hollow victory in fight to bring transparency to cops’ use of facial recognition technology
Connecting state and local government leaders
Police used facial recognition software to go ‘window-shopping for a suspect,’ says a New Jersey man arrested after facial recognition software found his image matched the suspect in video footage of a robbery.
This story was originally published by the New Jersey Monitor.
Francisco Arteaga was incarcerated, waiting to appear for a court hearing last fall, when he spotted a huge guy eyeballing him from the other side of the courthouse holding cell.
“This guy’s arms are like this, right?” Arteaga said, tracing imaginary Popeye biceps in the air. “He got no neck. He has a bald, shiny head. He’s looking at me with this mean face. I’m like, ‘Oh, my God!’ He’s walking towards me. He goes, ‘Your name Arteaga?’ I said, ‘Yeah.’ He opened his big arms, and he hugged me. He goes, ‘Thank you, thank you so much! Because of your case, I’m going home!’”
It’s been 14 months since the court ruling that made Arteaga famous, at least among civil rights advocates, New Jersey defense attorneys, and defendants who have found themselves in legal trouble because of facial recognition technology.
Police relied on the technology to identify Arteaga as their prime suspect in the 2019 armed robbery of a Hudson County cell phone store. He denied any involvement, police had scant other evidence, and Arteaga, a Queens native, said he’d never even been to New Jersey. But authorities charged him anyway because facial recognition software spit out his mugshot as a match with grainy footage of the robber caught by surveillance cameras.
Arteaga challenged his arrest and demanded detailed information about the technology police relied on to identify a suspect, aiming to expose its flaws and exonerate himself. He won, with a state appellate panel ruling last year that he deserved to get those materials through discovery.
The decision might have been a watershed moment for criminal justice reformers, offering hope for defendants like Arteaga and his big, bald cellmate who have been charged in otherwise flimsy cases because of such digital deductions.
But at least in Arteaga’s case, prosecutors said they couldn’t provide details about the facial recognition technology that led to his charges, largely because the match was made in another state — outside their jurisdiction and the New Jersey appellate court’s reach.
By then, Arteaga had been behind bars as a pretrial detainee for nearly four years. Rather than remain in prison to continue his fight, he pleaded guilty, his mind on his young son, his teenage daughter, and his fiancee.
“I’m like, do I want to roll the dice knowing that I have children out there? As a father, I see my children hurting. I’m hurting, but I could hurt, right? I could deal with that. But when I see my hurting is affecting my children, I got to be a father. I got to go home to my kid,” he said.
Arteaga’s experience exposes gaps in regulation and oversight that are growing as law enforcement agencies increasingly turn to new technological tools to crack cases where traditional sleuthing has failed, said Dillon Reisman, a staff attorney at the American Civil Liberties Union of New Jersey. Reisman specializes in surveillance, artificial intelligence, and other new technologies.
Policymakers’ inaction to close those gaps puts everyone at risk of wrongful arrest and prosecution, Reisman added.
“What this case really warned us about is the threat of unchecked surveillance power and the acquisition of all of this surveillance technology with no accompanying accountability framework,” Reisman said. “All of these systems, by their nature, involve interstate cooperation and systems that are bigger than one individual agency, and that makes it extremely difficult to have any sort of transparency, to learn how the system’s used, to learn how the system might be flawed, and to advocate against it.”
‘Window-shopping for a Suspect’
The appellate ruling lays out how Arteaga wound up in the crosshairs of law enforcement.
The day after Thanksgiving in 2019, a gunman held up the Buenavista Multiservices store on Bergenline Avenue in West New York, pistol-whipping an employee and escaping with $8,950. The employee described the robber as a “Hispanic male wearing a black skully hat.”
West New York officers submitted images from store and area surveillance cameras for facial recognition analysis to the New Jersey Regional Operations Intelligence Center, a division of the state police. An investigator there found no matches but offered to repeat the inquiry if detectives produced a better image.
Instead, detectives sent the raw footage to the New York Police Department’s Real Time Crime Center. A detective there captured several still images, compared them against the center’s databases, and identified Arteaga’s mugshot in December 2019 as a “possible match.” Two store employees — including one who wasn’t at the store at the time of the robbery — confirmed Arteaga from a photo array as the robber. Arteaga’s mugshot was in the NYPD’s system from two convictions years earlier on non-robbery offenses in New York.
Detectives’ decision to farm the case out to the NYPD showed they went “window-shopping for a suspect,” Arteaga said.
“The police were like, ‘Well, you know what, let’s send it to New York with no documented reason to do so. We’re going to abandon our state’s professionals and we’re going to go to another state and start looking in their pool right now,” Arteaga said.
Arteaga had an alibi. He told police he was visiting relatives in Croton-on-Hudson in Westchester County the day of the robbery, and his aunt’s friend saw him there. But the friend died of COVID-19 before he could vouch for him, Arteaga said. Police charged him, and a judge ordered him held at the Hudson County jail until his trial.
His defense attorney filed a motion seeking information about the facial recognition software that identified Arteaga, including its name, manufacturer, algorithms, error rates and source code, as well as the qualifications of the analyst who ran the search, details about the mugshot database where the analyst got Arteaga’s photo, and any alterations the analyst made on surveillance stills to improve the odds of a match.
The trial judge denied the motion in May 2022, and Arteaga appealed. A three-judge appellate panel in June 2023 sided with Arteaga and returned the case to trial court, directing the judge to order prosecutors to provide the information the defense sought.
“Here, the items sought by the defense have a direct link to testing FRT’s reliability and bear on defendant’s guilt or innocence. Given FRT’s novelty, no one, including us, can reasonably conclude without the discovery whether the evidence is exculpatory or ‘merely potentially useful evidence,’” Judge Hany Mawla wrote.
Courts must work to understand new technology and allow the defense a meaningful opportunity to fully examine it, Mawla wrote, citing a 2021 state appellate ruling.
“Defendant must have the tools to impeach the State’s case and sow reasonable doubt,” he wrote.
Prosecutors did not appeal the ruling, which means it now carries the weight of law.
Attorneys at the state public defender’s office who represent defendants identified by facial recognition software say they’ve encountered the same problem Arteaga had — prosecutors insisting, despite their constitutional duty to provide exculpatory evidence to the defense, that they have no information on the technology underlying their cases, said Tamar Lerer, who heads the office’s forensic science unit.
“It’s a due process violation not to provide it,” Lerer said. “The state is still using facial recognition and I know that attorneys are not being provided with this discovery, and when they’re asking for it, they’re told that they can subpoena themselves. So we have a systemic problem with the lack of compliance with this decision.”
Adding to the complexity, some facial recognition developers require customers to sign non-disclosure agreements to protect their products from competitors. Such secrecy has sparked several lawsuits against the New York Police Department, which was ordered in 2022 to release records — and they showed the department has used Clearview AI, a controversial facial recognition technology former New Jersey Attorney General Gurbir Grewal banned in 2020.
“The police departments are very well aware that they’re utilizing tools of secrecy,” Arteaga said. “So when they take the software recommendation that is built on secrecy to target somebody, that person that has been targeted is sh*t out of luck.”
Lerer said her office is “waiting to see what’s next.”
“The defense is not supposed to be a regulatory agency,” she said.
The West New York store’s owner didn’t respond to the New Jersey Monitor’s request for comment, and a spokeswoman for the Hudson County Prosecutor’s Office declined to comment.
In February 2022, the New Jersey Attorney General’s office began soliciting public input about facial recognition technology to help shape a statewide policy on its use by law enforcement agencies. No action has been taken since the public comment solicitation, said Michael Symons, a spokesman for the Attorney General Matt Platkin’s office.
The office doesn’t track how many agencies in New Jersey use the technology, Symons added.
Reisman said the need for policymakers to act is becoming increasingly urgent as the industry expands.
“Facial recognition has been an area of computer science research for well over 30 years, but in the past decade, what we’ve seen is an explosion in the number of companies offering these services and in federal funding for local and state governments to acquire these systems. There’s a lot of money being thrown at these tools without a lot of accompanying forethought into the sort of controls and safeguards we need,” Reisman said. “It’s a terrifying place for the state to be.”
Arteaga considers himself a “hostage” for the time he spent behind bars, where conditions can be notoriously abysmal. Since he got out in November, he has worked to rebuild his life.
He’s studying holistic medicine online, sells health and nutrition products, and works as a gym teacher at a senior citizens’ center in Queens. He now lives in Union City to better accommodate his monthly meetings with his parole agent and sees his son and fiancee, who still live in Queens, every few weeks.
While he won his appeal, he said, it felt a bit like winning the battle but losing the war.
“People be like, ‘Yo, you had a good thing with your case. Why didn’t you fight all the way?’” Arteaga said. “My question for them was, ‘What would you do if you was me?’”
He hopes someone else will pick up the fight.
“I set up the putt close enough for somebody else to sink it in,” he said.
NEXT STORY: Rising Stars nominations extended until August 30