An NYPD command center (photo: Edward Reed/Mayor's Office)

In the four months since the NYPD first published protocols limiting its use of facial recognition software, the coronavirus-caused state of emergency and recent civil unrest over police brutality made the considerable leeway in the policy all but total. Absent any formal legal framework at the city, state or federal level, the internal, non-binding departmental protocols are the only public policy governing New York's lawless frontier of biometric surveillance.

The police department has insisted it has not taken advantage of the expanded leeway, while privacy advocates are concerned about the potential for abuse. State lawmakers have introduced legislation that would curtail the use of facial recognition and other biometric surveillance, but its fate is unclear. There is no city legislation despite the City Council's recent effort to pass police reforms, including a new law that will mandate the NYPD disclose the surveillance tools it employs.

Facial recognition programs use artificial intelligence to match photographs with faces from digital databases, helping authorities identify anyone, from a figure in a crowd to the pedestrian who believes they are anonymous on the sidewalks of New York City.

The NYPD has 20,000 security cameras, a fleet of surveillance drones (equipped with powerful zoom and thermal imaging), and 24,000 officers roving the streets with body-worn cameras, all the images from which could be analyzed with facial recognition software.

Facial recognition is still a relatively nascient technology that has been shown to be less accurate on members of non-white minority groups, gender nonconforming people, and others. Privacy advocates believe once the technology is perfected the NYPD will, through its Domain Awareness System -- the department's bank of surveillance footage and data -- be capable of tracking a target's every move in public.

The new NYPD policy was published in the department's internal patrol guide on March 12, only after a BuzzFeed News investigative report revealed the NYPD was the single largest client of Clearview AI, the controversial company that developed one of the world's largest face databases by combing the internet for billions of images.

"This new policy clearly defines the permissible use of facial recognition technology and it strikes the right balance between public safety and privacy," said Police Commissioner Dermot Shea in a statement when the policy was adopted, two weeks after the BuzzFeed report.

Under the policy, officers may scan footage for faces if, among other reasons, they believe a crime has been committed or is about to be committed, or if there is "an imminent threat to health or public safety." Facial recognition can also be used in missing persons cases, to identify potential crime victims or witnesses, or people under arrest who refuse to identify themselves. There are no expressly prohibited uses.

Experts and privacy advocates warn that latitude makes just about anything subject to artificial intelligence surveillance without a warrant, from marching without a permit to simply looking suspicious.

"The protocols described by the facial identification policy are quite broad and are still open to rampant abuse," said Albert Fox Cahn, founder and executive director of the Surveillance Technology Oversight Project (STOP) at the Urban Justice Center.

Under the state of emergency, where social distancing orders make simply being in public a potential crime, the number of activities that can constitute a violation are far greater. Governor Andrew Cuomo's executive orders restricting public activity because of the COVID-19 pandemic, which happened to go into effect within days of the NYPD's facial recognition policy, give local governments the authority to issue civil and criminal penalties for violations, such as being within six feet of others without a face covering.

The NYPD has been enforcing social distancing throughout the state of emergency, at times using helicopters and drones to identify potential violators. Between March 16 and May 5, police arrested 120 people in connection with social distancing violations, more than two-thirds of whom were Black, and a quarter of whom were Hispanic.

"I think there is always risk that the NYPD is going to use 'imminent' risk and 'threat to health or safety' as a catchall authorizing any number of uses," said Michael Sisitzky, a New York Civil Liberties Union lawyer who specializes in police transparency. "So the policy is not anywhere near as limited as it would seem on its face."

"These policies are crafted to give the NYPD carte blanche to use tools and technologies as they see fit," he added.

Advocates want to see legal limits set on the use of surveillance technology and protections against its abuse by law enforcement.

Since March, much public activity has been illegal, and simply being outside at a certain time could be the basis of criminal suspicion. Since the end of May, when massive protests for police reform, as well as some riots and looting, erupted, being any one of the countless New Yorkers found outside could mark a person for enhanced surveillance. That risk only expanded when the early-June curfew meant being on the street at night was explicitly outlawed.

Officials, including Mayors Bill de Blasio and Michael Bloomberg, have said facial recognition is an important tool to combat terrorism and part of a necessary trade-off among safety, privacy, and the First Amendment.

“As the world gets more dangerous, people are willing to have infringements on their personal freedoms that they would not before,” Bloomberg said of the emerging technology in an interview with the Associated Press in 2011.

"We’re the number one terror target in America, and we need exceptional tools to fight terrorism, de Blasio echoed eight years later, in June 2019, on NY1 when asked about the civil rights concerns associated with facial recognition. "My mission is to keep everyone safe while respecting civil liberties, so it’s something we’ll have more to say on, but it’s not an either-or in my book."

In the NYPD’s Public Security Privacy Guidelines, last updated in 2009, it says the Domain Awareness System will monitor only public places "where no legally protected reasonable expectation of privacy exists."

The police department has used facial recognition for years in relation to criminal activity other than terrorism, at least as far back as 2012, when the tech helped identify a suspect in a barbershop shooting.

In 2019, before the patrol guide was updated, the Facial Identification Section -- the NYPD's designated unit of investigators who run and assess requests for AI matches -- received 9,850 inquiries and identified 2,510 possible matches, according to the department's website. Roughly half of the matches were related to murder, rape, felony assault, robbery, or grand larceny, according to the site, but it’s not clear how many are crime suspects.

The mayor's office and police department did not respond directly to Gotham Gazette inquiries about whether participating in recent demonstrations, violating the curfew, or breaking emergency executive orders were considered imminent threats.

"The NYPD uses facial recognition as a limited investigative tool, comparing a still image from a surveillance video to a pool of lawfully possessed arrest photos," wrote NYPD spokesperson Al Baker in an email. "This technology helps bring justice to victims of crimes. Any facial recognition match is solely an investigative lead and not probable cause for arrest – no enforcement action is ever taken solely on the basis of a facial recognition match."

The mayor's office did not respond to requests for information or comment for this story.

A police department spokesperson told Gotham Gazette its Facial Identification Section, which is primarily responsible for assessing computer matches, has not been involved in enforcing emergency executive orders, like social distancing, and has no current requests related to recent demonstrations or the curfew, which lasted from June 1 to 7. (The NYPD did not say whether FIS had received requests related to the demonstrations at the height of activity over a month ago.)

"I guess NYPD is asking us to take their word for it," said State Senator Brad Hoylman, who sponsors a bill that would ban biometric surveillance by law enforcement statewide. "As it stands now we have no substantive oversight of their use of facial recognition."

The NYPD did not respond when asked whether members of the department outside FIS are authorized to use facial recognition technology.

"As far as I'm concerned, the Patrol Guide is kinda garbage on this point," said Cahn, of STOP. The policy seems to be limited to the activities of FIS, he said, "but that doesn't mean it's a comprehensive set of regulations for all forms of facial recognition."

BuzzFeed's Clearview AI report notes over 11,000 searches were made by more than 30 officers, all before the patrol guide policies were put in place. The NYPD website now says facial comparisons are limited to the department's mugshot database, but certain top brass still have the discretion to scan photos against outside databases.

There is no guarantee that images captured at the protests or any other public activity will not be analyzed later using artificial intelligence like facial recognition software. The NYPD did not answer questions about how long it stores footage.

Most other elements of how the department uses surveillance technology are also under wraps. Advocates believe a new law called the Public Oversight of Surveillance Technology Act, or POST Act, recently passed by the City Council and signed by de Blasio, will shed unprecedented light on what equipment the department has by requiring the NYPD to report on the tools it deploys and rules for their use.

The jury is still out on whether a facial recognition match can be the basis of a criminal prosecution, but official responses from the police department and district attorney offices indicate that a computer match alone is not sufficient.

"Facial recognition is mostly used in the investigation stage," wrote Oren Yaniv, a spokesperson for Brooklyn District Attorney Eric Gonzalez, adding that it is not used as evidence at trials.

"Our office does not disclose law enforcement operations or investigative techniques," wrote Danny Frost, a spokesperson for Manhattan District Attorney Cy Vance, in response to questions about the office's facial recognition policy.

Civil rights observers warn that facial recognition can discourage or even be used to intimidate people from engaging in constitutionally protected activities like free speech.

First Amendment activities, like public demonstrations and participating in a political group, are supposed to be protected from warrantless police surveillance under a 1985 NYPD consent decree, known as the Handschu Agreement, adopted following a class action suit charging the police department with years of systematic spying on political organizations like the Black Panther Party.

"All the different ways that the police can monitor protests, both ways that are visible and invisible, can have a chilling effect," said Angel Diaz, a national security lawyer at the Brennan Center for Justice.

"There are people for whom the fact that there might be an ongoing police presence or that you might attract police attention may dissuade them from participating in a protest," he said.

This is especially true of people who are undocumented, or who have had past criminal justice involvement, in some cases the same communittees effected by the policing policies at the center of recent demonstrations.

"They might feel less comfortable participating because of the downstream consequences that can go on," Diaz said. "Especially when we don't know how long the data is being kept."

Nevertheless, surveillance equipment was out in force at the recent demonstrations against police misconduct. On Friday, May 29, the first evening with serious clashes between police and protesters, traffic light cameras -- installed to catch motorists running red lights -- flashed repeatedly above peaceful marchers gathered near Fort Greene Park in Brooklyn. On June 1, the first night of curfew, drones flew over marchers throughout the night. Through the weeks of protest, thousands of officers wearing body cameras have been deployed and police helicopters have been a fixture above marches.

It's not clear that all or any of the drones were the NYPD's and the department wouldn't say what types of surveillance tools were used on protesters. In 2018, then Police Commissioner James O'Neill announced the department had acquired 14 military-grade surveillance drones for special operations. Police reporting from the first quarter of the year, which doesn't include the recent demonstrations, looting, or curfew, show 20 drone flights took place between January and March, mostly for training and testing purposes.

The patrol guide says NYPD drones are not themselves equipped with facial recognition software and that their footage will not be subject to digital analysis, "absent a public safety concern."

Cahn says the disclaimer is vaguely worded and, thus, toothless because footage can later be processed with facial recognition. "Any video of a photo source can effectively become a facial recognition camera simply with that extra step," he said.

"In some ways, the Handschu agreement feels increasingly out of date given the modern way that police are doing surveillance," Diaz said. It comes from an era when policecraft was more physical, he said, focusing more on undercover officers infiltrating political movements than ubiquitous digital information gathering. "It doesn't specifically address issues raised by things like facial recognition or even the way in which modern social media monitoring works."

Lawmakers in Albany are carrying two bills that could change how surveillance technology is used by law enforcement. One, sponsored by State Senator Jessica Ramos and Assembly Member Ron Kim, two Queens Democrats, would prohibit the use of drones during activities protected by the First Amendment, and would require a warrant to use drones in criminal investigations. The law would also prohibit facial recognition from being used on drone footage, and sets up provisions for destroying the footage after a period of time.

Another bill, carried by Senator Hoylman and Assemblymember Deborah Glick, both Manhattan Democrats, would entirely prohibit the use of biometric surveillance by law enforcement and would create a task force to propose standards for its use in the future. The bill would ban biometric tools beyond facial recognition, like voice and gait analysis, but leaves carve-outs for more traditional police tools like fingerprints and DNA analysis.

"It's time to hit the pause button on this technology, gather experts, and develop some standards that require the NYPD and all law enforcement to follow," Hoylman said in an interview.

The police department has been resistant to any proposed limits on its use of facial recognition, and of surveillance equipment more broadly.

"I’d absolutely be opposed to a ban," said then-Police Commissioner James O'Neill at a press conference last October. "There is so much video in New York City today that to not use facial recognition would be irresponsible."