The Milwaukee Police Administration Building downtown. (Photo | Isiah Holmes)

Back in July, a lot happened while the Republican National Convention (RNC) was going on in downtown Milwaukee. Donald Trump accepted his party’s presidential nomination. Local residents protested the RNC. Out-of-state police killed an unhoused man in King Park, and the convention brought so much traffic to the gay and bisexual dating app Grindr that it crashed. Those events and more were probably followed by the Milwaukee Police Department (MPD) using a new tool to scan, scrape and search online activity. 

In April, the MPD announced that it was seeking an open source intelligence tool ahead of the RNC. Basically, anything which can be openly seen and accessed online counts as open source intelligence. Using the tool, the MPD planned to augment its online monitoring capabilities. What would have taken hours just a few years ago could be reduced to minutes. By the end of May, MPD had settled on an Artificial Intelligence (AI) powered software called Babel Street. The contract for Babel Street, which was not to exceed $43,673.50, was awarded on May 23.

A Request for Proposal (RFP) document, compiled by the Pewaukee-based technology brokerage company Abaxent, provides details on Babel Street. The document was obtained by Wisconsin Examiner through open records requests. Utilized by the U.S. Armed Forces, intelligence agencies and the federal government, Babel Street “empowers users to extend their search to the farthest corners of the globe, netting data beyond the traditional scope of [publicly available information] in a safe and secure environment,” the RFP document states. “It opens the door to enriched and standardized [publicly available information] data from over 220 countries.”

Not only can Babel Street search online content in over 200 languages, it also employs “sentiment scoring” in over 50 languages. A Babel Street glossary of terms webpage states that sentiment analysis involves determining “if a given text is expressing a positive, negative sentiment or no particular sentiment (neutral).” The RFP document also claims that Babel Street’s use of AI “accelerates investigations and uncovers connections.”  

An MPD spokesperson echoed that point, saying in an emailed statement to Wisconsin Examiner that the software has “increased the speed of investigations.” The spokesperson said that Babel Street is used by MPD’s Fusion Division. Social media investigations are a staple for Milwaukee’s Fusion Center, composed of both MPD’s Fusion Division and the Southeastern Threat Analysis Center (STAC). Originally created for homeland security, the Fusion Center serves a variety of roles today — whether that’s operating the city’s Shotspotter gunshot surveillance system, monitoring a camera network spanning Milwaukee County, conducting ballistic tests, accessing phones seized by officers, or processing information from cell towers. 

Within the Fusion Center, analysts assigned to the Virtual Investigations Unit monitor social media, investigating not only people but entire social ecosystems. Babel Street “pinpoints key online influencers, allowing investigators to explore networks from a powerful starting point,” the RFP document states. “Rapidly exposing and unlocking their web of relationships delivers crucial information in a matter of minutes.” All of that data then gets plugged into sophisticated visualizations such as maps, algorithmic scores, or graphs. “Visualized mapping unearths influencers who have the greatest impacts on organizations, senior leaders, and world events,” the document explains. “Advanced algorithms score and prioritize critical online entities to measure this influence, bringing to the forefront obscure identities that make up their network.” 

Babel Street can track the growth of online influence emanating from a person or group of interest to police. Investigators can also set real-time updates alerting them to new developments online, as well as “persistent” monitoring. “A persistent Document Search on an identified threat actor continuously monitors filtered topics the actor is publicly engaging in,” according to the RFP document. “By establishing a persistent collection via user-built filters/queries, users can not only increase their data access and insight, but they can also automate the rate aspects of analysis.”

Records from the City of Milwaukee Purchasing Division, obtained through open records requests.

Babel Street draws on a wealth of online information to gather intelligence for police. An aspect of the software known as “Synthesis” allows MPD “to understand the profile of key influencers based on attributes, such as person/organization, location, occupation, interests, areas of influence, and communication style, which are automatically tagged for millions of accounts using an AI model, while still giving the City the option of manual tagging.” Babel Street also allows MPD to pair keyword searches with geo-fencing, thus alerting the department to posts within a specific geographic area. MPD’s new open source intelligence tool also enables data to be extracted from the dark web — parts of the internet which are not indexed in search engines and require specialized internet browsers to locate.

The ability of law enforcement to map online connections between people worried privacy advocates leading up to the RNC. In early April, the American Civil Liberties Union (ACLU) of Wisconsin and Electronic Frontier Foundation (EFF) warned that using an open source intelligence tool, MPD could more effectively track and profile people who were exercising their constitutional rights. David Maass, director for investigations at the EFF, told Wisconsin Examiner that open source intelligence tools “are designed to produce ‘results’ even if there’s no evidence of a nefarious plot.” 

Many police reform activists in Milwaukee also remember the protests of 2020, when police departments heavily relied on social media to surveil protesters. All of that information, however, takes time to collect and sift, especially when a department may only have so many analysts on hand. “No longer are analysts manually checking multiple data sources to identify changes,” according to  the RFP document, “as Babel Street Insights persistently and automatically collects, ingests, and alerts users when new information is available, dramatically increasing the efficiency and effectiveness of each analyst.” 

All of that information, however, also needs to be vetted to ensure that it’s accurate. “Intelligence often requires vetting in order to determine whether it is reliable or not,” MPD’s spokesperson wrote in an email statement. “Additional investigation would be required with all intelligence.”

MPD said that it does not track Babel Street’s involvement in investigations, either during the RNC or after. There is also no standard operating procedure governing the software’s use by MPD, a spokesperson told Wisconsin Examiner. “This software is utilized to investigate crimes or to assist with mitigating threats to pre-planned large-scale events,” wrote the spokesperson in a statement. No decisions have been made yet about renewing the MPD’s one-year contract for Babel Street.