The Los Angeles police department worked with a Polish firm that specializes in “strategic communications” to monitor social media and collect millions of tweets last year, including thousands related to Black Lives Matter and “defund the police”, according to records reviewed by the Guardian.
Internal LAPD documents, obtained by the Brennan Center for Justice through public records requests, reveal that the department conducted a one-month trial of social media monitoring software from Edge NPD, a company that typically worked in advertising and marketing, had no prior experience contracting with law enforcement and was based thousands of miles away in Warsaw, Poland.
During the trial in fall 2020, Edge NPD tracked tweets on roughly 200 keywords for LAPD, the records show. In the process, the software collected millions of tweets, according to Edge NPD’s CEO, Dobromir Cias. The data set included tens of thousands of tweets related to Black Lives Matter and racial justice protests, some of them from prominent Black activists outside of LA and private civilians advocating for reforms, the files show.
The records suggest that LAPD was interested in using the company’s services, in part to help the department respond to “negative narratives”. Cias told the Guardian the company also aimed to flag possible threats.
The documents did not reveal what LAPD did with the data that was collected, and the department did not respond to repeated requests for comment.
LAPD ultimately did not pursue a permanent contract with the firm. But still, experts said, the trial raised serious concerns.
They wondered about the effectiveness of pulling in so much data, in so little time. Some said that, although law enforcement, journalists and researchers regularly monitor public online activity, it was troubling that a police agency’s social media monitoring activities appeared to include tracking a broad swath of critics. And the partnership also raised questions about oversight of surveillance technology, as well as police agencies’ data collection practices.
LAPD’s test run with Edge NPD came as law enforcement agencies have increasingly been forced to take their investigations online, and have sought tools to do so. Tech firms have responded by pushing new innovations and pursuing police contracts, and LAPD and the New York police department, with some of the largest budgets in the US, have often been at the forefront of piloting software.
Identifying ‘negative narratives’
Edge NPD primarily assists private companies with market research and helps them ensure that advertising campaigns aren’t jeopardized by bots and trolls, Cias told the Guardian.
The company was connected to LAPD by a US government agency that had used the firm’s software. A representative of that federal department emailed Edge NPD in September 2020, saying LAPD was interested in using its services for “public safety and strategic communications” and to “identify disruptive social media activity being artificially amplified by malign actors”.
ABTShield, Edge NPD’s proprietary software, could help LAPD “identify as early as possible when activity that could lead to civil unrest is being amplified via social media”, the US representative wrote, adding, “From a messaging perspective, knowing what the negative narratives being artificially amplified are would allow the communications team to create effective and timely responses.”
During the 40-day trial in October and November of 2020, Edge NPD provided LAPD with a dashboard monitoring tweets related to six topics: “civil unrest”, “American policing”, “domestic extremism and white nationalism”, “election security”, “potential danger” and the conflict between Armenia and Azerbaijan (which at the time was prompting local LA protests).
The two entities discussed roughly 200 possible keywords to follow, including “lapdchiefmoore”, “abolish the police”, “nojusticenopeace”, “police budget”, “police killing” and “acab” (a protest slogan that stands for “all cops are bastards”). In one email, Cias suggested adding “defund the police” and “BLM” to the list of keywords to follow.
An LAPD official agreed that BLM would be good to monitor, but added that “there are MANY legitimate people who are using that to express their rights”, records show.
Cias said that in addition to the dashboard, on occasion he would also personally send tweets to LAPD that he thought might signal a public safety threat.
“The major goal was to actually pass [along] quickly anything that looks kind of dangerous,” he said. He acknowledged he didn’t always confirm whether the content he was forwarding was legitimate: “When you’re passing this information, you don’t really know how serious it is. I think it’s up to law enforcement to really verify if it’s true … We don’t do fact checking.”
At one point, the records show, Cias emailed the department a post from a critic who had tweeted a video that appeared to show an LAPD officer tackling a journalist with a comment the department was “overfunded … and high on power”. (Cias told the Guardian he was sharing it as an example of how “defund” narratives were gaining traction online.)
The Brennan Center, which analyzed the tweets, found that the service collected nearly 2m tweets during the trial, including roughly 270,000 posts under the “American policing” category.
The records show the software frequently captured tweets with mainstream news articles and private civilians expressing opinions. The system also flagged tweets from LAPD critics calling on the chief to “resign”, as well as celebrity posts, including a tweet from Common, the rapper and activist, about racism in America.
‘It creates a haystack’
Experts consulted by the Guardian raised questions about the trial’s set-up, doubting the effectiveness of collecting such large amounts of data.
Mary Pat Dwyer, a Brennan Center fellow, questioned why LAPD was wasting limited resources to “chase content online that doesn’t pose any kind of safety threat”, including ordinary political speech and criticisms of police. “It’s striking the volume of information that they were pulling in and the terms they were using. It’s hard to understand how LAPD would even be processing all of this.”
“It creates a much larger haystack of data that doesn’t actually lead to any real, tangible, positive outcome for communities,” said Steven Renderos, executive director of national racial justice organization Media Justice. “Instead it just equips the police department with more data that helps it kind of justify its own efforts to to tell a better story about itself.”
He pointed out that the list included keywords such as “domestic extremism” and “racist” that were purportedly intended to capture tweets about white supremacy. “No one is describing themselves as a domestic extremist,” Renderos said. “Instead you then start filtering in tweets and messages from people who are merely … protesting white supremacy. Take these keywords together and what it’s actually doing is capturing a subset of thought and dissent among people. And that’s dangerous.”
They also worried what an agency like LAPD could ultimately do with such data. “I don’t trust the Los Angeles police department to use a tool with this amount of data in a way that’s responsible, because history has shown us that they can’t,” said Renderos.
LAPD has faced several racial profiling scandals over the years. The Guardian also recently revealed that LAPD was engaging in broad collection of civilians’ social media data, and had partnered with a different tech firm that claimed its algorithms could identify people who may commit crimes in the future, with criteria that experts said was discriminatory. Those revelations prompted Facebook to demand that LAPD stop collecting data on its platform for surveillance.
Activists caught up in the surveillance said they were not surprised. The software flagged tweets by Bree Newsome Bass, who received national attention in 2015 when she climbed a flagpole to remove a confederate flag at the South Carolina statehouse. In a thread included in the files, Newsome Bass advocated for demilitarizing police and putting funding toward mental health first responders. “The resistance to defunding police is 100% about racism & normalizing the daily violence police forces inflict on marginalized communities for the benefit of ruling classes,” she wrote.
“It’s political targeting,” Newsome Bass told the Guardian. “We’ve seen instance after instance where police agencies are focused more on policing Black people who are demanding equality and civil rights than actually preventing any violence … They’re making the case for defunding the police even further. They’re using taxpayer dollars to monitor our social media where we’re talking about how we’re wasting money on police.”
“There’s nothing violent or criminal about saying ‘defund the police’,” added Melina Abdullah, co-founder of Black Lives Matter LA. “We have a right to say ‘defund the police’. The idea that LA is tracking this kind of ideology should be very scary to people.”
Combatting ‘defund the police’
Cias told the Guardian the trial was free for LAPD and meant to be a “demonstration” of its technology, and to “help LAPD detect potentially dangerous situations” during the tense election period. He said the service identified roughly a dozen alerts related to possible threats and provided examples of tweets about protest clashes and about Dodgers fans allegedly engaging in vandalism while celebrating.
He stressed his firm is not a defense contractor, and the service was not intended to monitor specific activists: “This was not for actually analyzing the members of the Black Lives Matter movement.”
In a thank you letter to the company after the trial, an LAPD official said ABTShield had surfaced “threat-related content” and “enabled us to more efficiently analyze this barrage of data”. The service helps “parse out the proverbial ‘signal from the noise’”, the official said. In another email, an LAPD representative wrote, “The product is definitely working.”
As Edge NPD was starting its trial, the company also proposed a $150,000 annual contract for LAPD. The company told LAPD it could use artificial intelligence, machine learning and “human analysis” to identify online narratives that were rooted in “disinformation”, as well as public safety threats.
The firm proposed LAPD monitor specific “client-identified” topics, including “LA riots”, “police violence”, “BLM protests” and the “Proud Boys”. It also suggested that LAPD was the direct subject of misinformation and “malicious activity”, writing: “The LAPD itself is being targeted by organized attacks of automated bots and trolls (e.g. police brutality misinformation and “defund the police” narratives).”
Asked for clarification, Cias told the Guardian he did not have specific evidence of bot attacks on LAPD or of trolls spreading “defund the police” misinformation, but that the discussion was based on a “very strong hypothesis”.
“I could qualitatively assume that it might be true that those narratives related with Black Lives Matter and defunding police might be also supported … to some extent by malign actors who are interested in actually disorganizing public institutions in the United States,” he said, adding that he was proposing, in effect, doing “market research”.
While there is evidence that Russian trolls amplified Black Lives Matter content during the 2016 US election, activists said they were concerned that police departments and other critics of their movement were conflating authentic organizing online and troll campaigns.
“There’s been a lot of grassroots organizing to get people involved in the budget process and to put a spotlight on the absurd amount of money we spend on policing,” said Kenneth Mejia, an LA housing justice activist and city comptroller candidate who advocated for defunding. He noted that last year, BLMLA facilitated a community-driven process to present a proposed “people’s budget” that cut LAPD funding. “There’s an organic and growing awareness of the reallocating of resources from the police.”
LAPD did not move forward with a formal contract with Edge, though has remained interested in this kind of service, with records showing the department purchased or pursued software from at least 10 companies that monitor social media.
Last week, Twitter said it had suspended ABTShield’s developer account based on evidence that the firm violated its policies by deviating from approved uses, though the company did not elaborate further on the decision. Twitter’s policies allow for public data to be used for “news alerting” and “first responder support”, but it prohibits surveillance of “sensitive groups”, such as activist organizations.
A Twitter spokesperson, Shaokyi Amdo, said in a statement: “Twitter prohibits the use of our developer services for surveillance purposes. Period. We proactively enforce our policies to ensure customers are in compliance and will continue to do so.”
Evan Greer, deputy director of Fight for the Future, a digital rights group, said there should be stronger restrictions against the broad collection of social media data and the analysis law enforcement may perform on it.
“People don’t fully grasp the ways in which the ability to analyze data at a mass scale changes the game. There’s a difference between you tweeting something and knowing that it may be seen in public, and you tweeting something and knowing that it can be vacuumed up and analyzed in a million different ways using artificial intelligence and machine learning.”