New Police Body Cam Data Exposes the True Scale of NYPD Violence Against Protesters

A landmark $13 million settlement with the City of New York is the latest in a string of legal wins for protesters who were helped by a video-analysis tool that smashes the “bad apple” myth.
Video: Cameron Getty; Angela Weiss/Mario Toma/Rise Images/Getty Images

Beaten, blinded by pepper spray, corralled like animals, and indiscriminately arrested for marching against police violence and racial injustice. Such was the fate hundreds of people suffered at the hands of New York Police Department (NYPD) officers in late May and early June of 2020, as thousands of people across the United States protested the murder of George Floyd by a Minneapolis police officer.

Three years later, the City of New York has agreed to pay $9,950 each to some 1,380 protesters as part of a settlement of a class action lawsuit. Costing taxpayers more than $13 million, it’s the largest amount paid to protesters in US history, according to the protesters' legal team.

Lawyers secured the settlement with the aid of a little-known tool that helped them quickly categorize and analyze terabytes of video footage from police body cams, helicopter surveillance, and social media. “We had multiple weeks of protests. We had protests spanning the city of New York. We had thousands of arrests,” says David Rankin, a partner at the law firm Beldock, Levine & Hoffman who was part of the protesters’ legal team. “We had tens of thousands of hours of body cam footage, we had text messages, we had emails, we had just an absolute truckload of data to get through.”

The path through all this data was carved by Codec, a video categorization tool developed by the civil liberties-focused design agency SITU Research. Launched in June 2022, the tool is proving essential in legal battles around the world, where hours of disparate video footage can reveal orchestrated, state-backed violence against protesters.

Clip by Clip

Dozens of videos shared with WIRED show how the legal team built their case. Using this data, which also included geospatial information, time stamps, and the category of the alleged misconduct, we were able to build a map that allows anyone to watch the police incidents that were central to the lawsuit. Each dot represents an incident the legal team characterized as police misconduct. Of the 72 videos the legal team flagged as most pertinent to their case, the map includes 47 videos recorded by police body cams or surveillance cameras. The locations of the remaining 25 videos, which appear to have been taken from social media and other sources, are also pinpointed on the map. In total, the legal team analyzed more than 6,300 videos. 

Some of the videos on the map contain graphic violence, and viewer discretion is advised. Videos will autoplay with the sound on.

Video evidence was obtained from various sources, including police body cameras and helicopters. SITU Research geolocated the data and provided it to WIRED.

Among the videos we reviewed, an NYPD officer can be seen running down the sidewalk while pepper-spraying a person who’s standing against a building, entirely out of the officer’s way. In another video, an officer hits a protester with a car door while driving down the street. Another video shows a group of officers interlocking arms as one of them says, “Just like we fucking practiced.” The officers then charge a group of protesters before singling out a person on the sidewalk and beating them with batons. Taken together, the footage demonstrates widespread, systematic police misconduct during protests that spanned from May 28 to June 4, 2020, across multiple neighborhoods in New York City, according to the lawsuit.

While looting and vandalism took place in several neighborhoods during the protests, the demonstrations were largely peaceful. The defendants in the lawsuit have not admitted wrongdoing as part of the settlement, and city attorneys deny an orchestrated effort to violate protesters’ rights. Reached for comment, the NYPD referred WIRED to the city’s Law Department, which has not yet responded to a request for comment.

Remy Green, a partner at Cohen & Green and a member of the protesters’ legal team, says the use of police body cameras, which have been touted as a step in the right direction for civil liberties, has become “a kind of band-aid solution to police brutality.” A single video can only reveal so much, Green says, and police departments can use this limitation to obfuscate what really happened. Protests that are met with an extreme police response require a zoomed-out vantage point, which is what Codec allowed the legal team to create. “It gives you a much more comprehensive look at the activities that occurred,” says Green.

Hard Pivot

The idea of using Codec in the lawsuit came from a related case settled earlier this year. Here, Human Rights Watch worked with SITU Research to analyze video footage of protests in the Mott Haven neighborhood of the Bronx, one of New York City’s five boroughs. Their work proved that the NYPD used an anti-protest tactic called “kettling”—trapping a group of people so they can’t escape—just prior to a government-mandated curfew, thus ensuring that they were in violation of the order. In March, a lawsuit against the city over the NYPD’s use of kettling ended in a $21,500 payout to each of more than 300 Mott Haven protesters, which is estimated to be the highest per-person settlement for a mass arrest in US history. The NYPD said in a statement following the settlement that it has since “re-envisioned” its “policies and training for policing large-scale demonstrations.”

Having seen the forensic video investigation SITU’s work on the Mott Haven protests produced, Rankin asked for help conducting a similar investigation. But this time it wouldn’t focus on police conduct during a single protest in one neighborhood, but rather protests across New York City.

Now an open source tool that’s free to use, Codec was first built for internal use only, according to Brad Samuels, a founding partner of SITU who oversees its research division. Codec categorizes video files using a range of criteria, including the location where a video was taken and the time and date it was recorded. Analysts can add their own parameters, such as whether a video shows police use of pepper spray, which enables them to more easily pinpoint clips that are relevant to an investigation.

Early on, the SITU team focused on international incidents, such as the 2014 Maidan Revolution in Ukraine and a 2019 uprising in Iraq in which security forces fired military-grade tear gas canisters directly at protesters’ heads. After the murder of George Floyd in May 2020, the US-based SITU team “pivoted hard” to investigate domestic incidents. The agency’s focus on the 2020 Black Lives Matter protests that followed resulted in investigations into police misconduct in Portland, Oregon, and Charlotte, North Carolina. Investigations in Mott Haven and across New York City are the latest to join that list.

The legal team behind the New York City-wide lawsuit, which operated through the National Lawyers Guild, a civil rights-focused nonprofit, obtained thousands of videos from the NYPD, the city, and social media feeds, most of which were unpublished until this week. Using Codec, analysts pinpointed four broad categories of what the legal team characterized as police misconduct: improper arrests, improper use of pepper spray, improper baton strikes, and excessive force.

While Codec greatly enhanced the legal team’s ability to analyze thousands of videos, it still took “a lot of human hours” to make the data usable and identify specific moments that were useful for their case, Rankin says. This included identifying the location of specific police interactions, categorizing the alleged misconduct, and adding accurate time stamps—all of which require ample hours of human work to complete, even with Codec’s help.

In future cases that require high-volume video analysis, however, that job may become easier. Samuels says the SITU team is currently working to add layers of machine learning and computer vision to Codec. He hopes that by next year the system can be tasked with, for example, detecting batons in footage—something that is currently done manually.

Widespread protests like those that took place in 2020 don’t often happen in the US. And Rankin points out that cases like the one settled this week are rare. But Green says they’re not nearly rare enough. “Our role as civil rights attorneys, I like to think, is to attempt to put ourselves out of a job,” Green says. “This settlement marks a high-water mark in terms of financial accountability, but I don’t see myself losing my job anytime soon.”

Map creation and additional data work by Danielle Carrick