On January 30, 2020, the Brennan Center submitted a request under the California Public Records Act (CPRA) to the Los Angeles Police Department (LAPD) for information on the department’s use of social media to collect information about individuals, groups, and activities.
Read the CPRA request here.
The LAPD acknowledged receipt of the Brennan Center’s request and sent a partial response on February 24, 2020, with two documents that had already been released publicly. On March 12, 2020, the LAPD produced one additional document and closed the request. These responses did not include materials that had already been made available in response to other public records requests or referred to in media articles, indicating that the LAPD’s search was deficient. On November 17, 2020, the Brennan Center filed a lawsuit in the Superior Court of California, County of Los Angeles, with assistance from our pro bono counsel, Davis Wright Tremaine LLP. The suit challenged the LAPD’s inadequate search and failure to produce responsive documents. In response, the LAPD released supplemental productions between March and September 2021.
The LAPD’s Productions
In total, the Brennan Center has received twelve sets of documents totaling over 10,000 pages.
The initial production consists of memos outlining guidelines for officers’ use of social media. A memo from October 2018 outlines procedures to create an official LAPD social media account and policies that govern officer conduct on social media for official business and personal use. These guidelines do not apply to approved undercover activity on social media. A memo from March 2015 introduces the LAPD’s Social Media User Guide, which provides direction on Department-sanctioned uses of social media to advance the LAPD’s community relations mission, develop situational awareness, and act as an investigative tool. The Guide also briefly addresses the First and Fourth Amendment implications of police officers’ use of social media to monitor or collect information.
Each series after the first production is labeled with consecutive letters. The A-Series consists of a memo from the LAPD’s Chief of Detectives dated April 27, 2018, directing concerned personnel to retain information gleaned from social media that “may disappear before, during, or after a crime.” The memo advises that retaining this information enables investigators to obtain a warrant for a social media account more easily. In a later production, the LAPD produced an instructional video on how officers can preserve information from social media; the video insinuates that African-American men who dress professionally in court are gang members “in real life,” and urges officers to set up a “dummy account” in order to access profiles on the platform.
The B-series includes three excerpts of emergency operations guides from 2009 and 2010 that direct officers to use social media tools to manage or gather information about “emergent” or “significant incidents.” The last page instructs the LAPD’s intelligence group to coordinate surveillance activities on social media, media information, and internet “chat and postings.”
The C-Series encompasses several documents, including a section of the Department Manual that describes the Computer Statistics Division (COMPSTAT), the LAPD division responsible for implementing, evaluating, and auditing the production and dissemination of all crime analysis products, including social media monitoring software– although the Brennan Center did not receive any audits of social media monitoring tools as part of the LAPD’s production. COMPSTAT is also responsible for providing training and support for Palantir software and preparing predictive policing reports weekly. (Palantir is a software company specializing in big-data analytics that developed and sells predictive policing software to local and federal law enforcement agencies.) The C-Series also includes a document setting out the Organization and Functions of the LAPD’s Community Relationship Division, an office within the LAPD responsible for social media monitoring until it was disbanded in the fall of 2018.
The D-series includes the Intelligence Guidelines for the Anti-Terrorism Intelligence Section of the LAPD’s Major Crimes Division. Among other guidelines, this document defines and distinguishes Online Investigative Activity (OIA) and Online Undercover Activity (OUA) for terrorism investigations. Though both involve the use of a “Fictitious Online Persona” as part of an investigation, OUA involves “on-going interactive communication […] with an identified person or group and is related to an ongoing Terrorism Intelligence Investigation,” while OIA appears to account for all other online investigative activity using a fictitious account. Furthermore, while officers engaged in OUA require departmental approval, OIA is mostly unregulated and does not require approval. The Guidelines specify that “Fictitious Online Personas created for the purposes of identifying and examining terrorist trends and tactics, developing profiles, or conducting threat assessments does not constitute Online Undercover Activity,” meaning that LAPD personnel can engage in a significant amount of undercover monitoring online without departmental oversight or accountability. The D-Series also includes a memo from former LAPD Chief Charlie Beck from May 2015 asking police officers to record subjects’ social media and email account information when completing field interview (FI) cards.
The E-Series includes emails between LAPD officials and companies that offer social media monitoring services dated between March 2016 and December 2020 – specifically, Geofeedia, Skopenow, and Dataminr. From March to May 2016, officers from the LAPD’s Real-Time Analysis and Critical Response Division (RACR) had a free trial with Dataminr, which was scheduled to coincide with a May Day protest. The LAPD had a demo with Skopenow in June 2019, and multiple trials between November 2018 and July 2020; Skopenow offered services including anonymous analysis of social media and access to the social media platform Parler. The documents also show that the LAPD licensed and used Geofeedia from 2014 to 2016, using funds from the privately funded LAPD Foundation. However, the LAPD stopped using Geofeedia in 2017 because Geofeedia lost access to Twitter and Facebook data after public records requests from the ACLU of Northern California showed that Geofeedia and similar companies were pitching their products as resources to surveil lawful protesters against police violence.
The F-Series includes emails between the LAPD and sales representatives for Babel Street, a service that surveils English and non-English language sources, dated between May 2016 and April 2020. The LAPD scheduled a demo with Babel Street in May 2016. This series also includes correspondence with DigitalStakeout as the LAPD prepared to apply for a grant for public safety technologies through the U.S. Department of Justice, although the LAPD ultimately did not pursue the grant. DigitalStakeout offers monitoring services using public data from the Internet, social media, and the dark web. There are also unsigned copies of the LAPD’s order forms for Geofeedia from January 2016, which include payment for Geofeedia’s “systems integration” with Palantir, marked as “to be discussed.”
The G-Series includes email correspondence between the LAPD and Cobwebs, an AI-powered internet monitoring service, dated between May 2019 and October 2020. An email from the LAPD to a Cobwebs sales representative from February 2020 states that the Robbery-Homicide division did not have the budget to license the Cobwebs platform. However, the officer indicated the LAPD was potentially interested in licensing Cobwebs in preparation for the 2028 Olympics. An email from October 2020 indicates that the LA District Attorney’s office was trying out Cobwebs. The same month, the LAPD’s Robbery-Homicide division received a quote for five Cobwebs licenses in preparation for its Urban Areas Security Initiative (UASI) grant request application.
The H-Series includes email correspondence from the LAPD with Dataminr and Media Sonar dated between December 2015 and January 2020. The LAPD trialed Dataminr services from March to June 2016 as well as in August 2019. The documents also reveal that the LAPD began receiving pricing proposals from Media Sonar in 2018 and purchased Media Sonar licenses in 2021. In a presentation it submitted as part of its proposal to the LAPD, Media Sonar sells itself as an “online investigation software” that enables users to “Detect potential threats, find new leads, locate witnesses, and improve [users’] crisis management.” The presentation also includes details of a “Digital Footprint” feature that builds “a full digital snapshot of an individual’s online presence including all related personas and connections.” Media Sonar claims to have access to over three hundred data sources with two billion records compiled from public records, “crowd-sourced data,” and more.
Since 2019, the LAPD has received order forms for two annual premium packages from Media Sonar, one for ten users and another for five. For the 2020 and 2021 fiscal years, the LAPD applied to purchase Media Sonar licenses through Urban Areas Security Initiative (UASI) grants from the Department of Homeland Security; these fund technologies that help “high threat, high density areas” prevent and protect against terrorism. According to the grant application, the LAPD would use Media Sonar for assistance “in the constant search for pre-event (terrorism) indicators, protective intelligence, and the efficient response/handling of critical incidents, whether criminal or terror based.” The LAPD’s UASI grant applications for 2020 and 2021 were approved. It is unclear whether the LAPD purchased and used Media Sonar licenses from 2018–2020, but in its response to the Brennan Center’s questions, the LAPD confirmed that it had purchased Media Sonar licenses in 2021 with funds from the UASI grant.
Furthermore, the grant application shows that the LAPD collaborates with the Joint Regional Intelligence Center, the area’s fusion center, in its use of open-source social media monitoring. The LAPD also produced an undated document with search terms that Geofeedia used for the LAPD’s Community Relationship Division, including #BlackLivesMatter, #BlackLivesMatter-LA, #SayHerName, Sandra Bland, Tamir Rice, #fuckdonaldtrump, and others.
The I-Series includes two documents relating to Field Interview (FI) cards. LAPD officers fill out FI cards to document people they have stopped or questioned; these cards can be completed on anyone an officer comes into contact with. The first document is a July 2020 memo from the Chief of Police, Michel Moore, to all LAPD personnel. In the memo, Chief Moore urges officers to diligently record all information in the FI cards, which would be subject to review by Department supervisors “for completeness and validity.” As noted above, former Police Chief Charlie Beck had sent officers a memo in May 2015 telling them to collect social media and email account information in FI cards. The second document is a copy of the FI card form, which shows that the LAPD is gathering subjects’ date of birth and social security number, with a disclaimer stating that subjects are obligated to provide their social security numbers upon an officer’s request. In their October 2020 audit of stops that the LAPD conducted in 2019, the Office of the Inspector General and the Los Angeles Police Commission recommended that the LAPD remove the field to collect social security numbers. Furthermore, the audit notes that the LAPD should revise its policies regarding the FI cards, so officers don’t complete them routinely as part of stops, especially after the LAPD ceased using the CalGang database in July 2020. The audit also emphasizes that members of the public are not under any obligation to answer LAPD officers’ questions to aid them in filling FI cards. The FI cards also have a field to collect social media and email account information. The Brennan Center surveyed other cities’ policies regarding FI cards and found no other police department that collects social media and email account information, though details are sparse.
The J-Series, released by the Brennan Center on November 17, 2021, includes email correspondence between the LAPD and Voyager Labs (Voyager) dated between October 2018 and March 2021. The documents reveal that the LAPD conducted a four-month trial of the company’s VoyagerAnalytics software between July and November 2019. Voyager touts its products’ “unique collection methods [that] enable traceless collection from social media networks,” which Voyager claims allow users to reconstruct closed profiles and closed Facebook groups based on publicly available information, as well as to uncover the strength and relationship type of people’s connections on social media. Voyager also claims its tools can identify people that are “most invested in a given stance: emotionally, ideologically, and personally.” The LAPD was particularly interested in the VoyagerGenesis tool, which analyzes warrant returns for social media information within minutes. Voyager also promoted its VoyagerCheck tool that automatically vets social media profiles based on a user’s list of predetermined questions.
To demonstrate these capabilities, Voyager sent the LAPD a case study from March 2020 titled “COVID-19 Outbreak: Investigating a Threat Actor.” Using the VoyagerAnalytics tool, the company analyzed the network of Bahgat Saber, a Muslim Brotherhood activist from New York City, after he told his followers to deliberately infect members of the Egyptian consulate with Covid-19. Voyager analyzed biographical information for Saber and his nearly 4,000 followers. Subsequently, Voyager analyzed Saber’s “indirect connections” (friends of friends), touting that it had found “affinity of at least some of Saber’s network for violent, radical ideologies.” Voyager also found that fifty-eight of Saber’s connections indicated that they lived in New York City. Of those living in New York City, Voyager classifies two people as threats because they work in U.S. government agencies, are associated with Saber, and have connections to “known extremist accounts.” In short, the case study presents a roadmap for guilt by association, classifying people as threats based on an artificial intelligence tool’s assessment of their direct and indirect connections on social media.
Voyager also sent the LAPD an undated white paper describing its capabilities with reference to the case of Adam Alsahli, who attacked the Corpus Christi Naval Air Station in May 2020. Voyager claimed that its tool can automatically vet and classify people by their risk of becoming Islamic fundamentalists. The results are color-coded (green, orange, and red) based on “artificial intelligence calibrations,” which Voyager describes as providing insights in minutes without human involvement. For Alsahli, Voyager highlights the fact that he posted “pictures with Islamic themes” that reflect “his pride in and identification with his Arab heritage” on Facebook and Instagram, and that AI analysis of these posts yielded an “orange” indicator.
In some correspondence with Voyager, the LAPD suggested that it did not want to use features that could be perceived as “active monitoring” of social media. In one instance, an LAPD representative asked Voyager not to activate a feature that would allow the LAPD to track changes in content and connections on a social media profile because it was too close to “active monitoring.” In another email, an LAPD representative expressed concerns about Voyager’s monitoring capabilities because of public outcry over law enforcement’s use of Geofeedia.
The series also includes several pricing proposals and price quotes that Voyager sent to the LAPD in 2019 and 2020. The LAPD and Voyager were discussing finalizing procurement, but the LAPD was unable to purchase VoyagerAnalytics licenses in 2020; the reason given was projected budget shortfalls during the Covid-19 pandemic.
This series also includes email correspondence from the LAPD with EDGE NPD dated between September 2020 and March 2021, which the Brennan Center released on December 15, 2021. EDGE NPD is a European software development company that created ABTShield, a social media monitoring tool that purportedly uncovers disinformation campaigns. In its promotional materials for the LAPD, EDGE NPD pitched ABTShield as a tool to help the LAPD at a time when “[the department] is being targeted by organized attacks of automated bots and trolls (e.g., police brutality misinformation and ‘defund the police’ narratives).” An LAPD representative put forward this narrative as well, asserting in an email to Dobromir Ciaś, the CEO of EDGE NPD, that there were “state-backed media outlets sowing division between law enforcement and the communities we protect.”
The LAPD conducted a trial of ABTShield in October and November 2020. As part of its trial, the LAPD received daily updates containing Twitter activity related to topics of interest the Department selected: domestic extremism and white nationalism, potential danger, civil unrest, election security, the conflict between Armenia and Azerbaijan, and American policing. To search through Twitter, the LAPD and EDGE NPD compiled an extensive list of keywords, which included words and phrases related to Black Lives Matter protests, white supremacist and domestic terror groups, the 2020 election, and more. The Brennan Center received a list with keywords that the LAPD requested, but EDGE NPD subsequently added additional terms and the LAPD agreed to the additions. According to the LAPD, it did not receive the final list of keywords used by ABTShield from EDGE NPD.
Finally, the K-Series, released by the Brennan Center on December 15, 2021, contained Excel files from the LAPD’s trial of ABTShield. On average, ABTShield sent the LAPD approximately 70,000 tweets every day (almost 2 million tweets in total) in three different spreadsheets: tweets with Los Angeles-specific content, tweets from Los Angeles (identified through Twitter’s geolocation tools), and English-language tweets. The spreadsheets include the content of the tweet, the originating account, and how many people viewed the tweet (i.e., the tweet’s reach). Additionally, ABTShield compiled over 1,400 tweets from social media handles that the LAPD requested ABTShield track; ABTShield tagged these tweets almost entirely as related to American policing and civil unrest. The Brennan Center’s review of the spreadsheet indicates that the LAPD tracked at least three accounts — two anti-fascist groups and one account that provides updates on protests based on information obtained through police scanners and civilian reports.
The LAPD’s Written Responses
Additionally, on April 28, 2021, the LAPD sent a letter to the Brennan Center accompanying its supplemental document production. In the letter, the LAPD represented that it had not conducted any audits regarding the Department’s use of social media monitoring. The Department also stated that records reflecting warnings or disciplinary action against LAPD employees for their use of social media are “confidential and non-disclosable,” and were therefore not produced.
On June 25, 2021, the LAPD sent a letter responding to the Brennan Center’s questions about the produced documents. The letter reveals that the LAPD obtained funding through a UASI grant in 2021 to procure Media Sonar licenses; however, the LAPD stated that it is not using other products listed in the UASI grant application, such as Skopenow, Cobwebs, ABTShield, and VoyagerAnalytics. The LAPD also represented that while the Department will be using Media Sonar in FY 2021, it has not signed a contract with the company. Although the LAPD’s September 2021 After-Action Report Implementation Plan states that the LAPD is using Voyager, the LAPD’s counsel believed this reference is an error.
On July 13, 2021, the LAPD sent a letter responding to the Brennan Center’s questions about the produced documents. The LAPD represented that it does not track how many field interview (FI) cards officers complete, nor the number of unique individuals about whom FI cards are completed.
On December 8, 2021, the LAPD responded to the Brennan Center’s questions about the keyword list used by ABTShield for the LAPD’s pilot. According to the LAPD, EDGE NPD did not provide a final version of the keyword list used in the trial to the Department.
