Skip Navigation

NYPD Predictive Policing Documents

Documents related to the NYPD’s predictive policing trials and programs.

Last Updated: August 6, 2021
Published: July 12, 2019

The Brennan Center filed a Freedom of Information Law (“FOIL”) request for information about how the New York Police Department (“NYPD”) has tested, purchased, and implemented predictive policing software on June 14, 2016. The documents below were ultimately obtained over the course of two years of litigation on that original request.

These documents reveal details about the NYPD’s test of three types of commercial predictive policing software and shed light on the in-house algorithms the NYPD ultimately employed to inform deployment decisions. The difficult process to get access to this information, and the piecemeal production we ultimately received, also reveal the NYPD’s quest to keep the public in the dark about this technology. The timeline available here arranges the documents the Brennan Center obtained in chronological order and summarizes their contents.

From these documents, the Brennan Center learned that, between January 2015 and June 2016, NYPD communicated extensively with three different commercial predictive policing providers, PredPol, Azavea, and Keystats. Between April and June 2016, the companies participated in a “Predictive Policing Pilot Evaluation,” a comparative trial run that tested the accuracy of their predictive software. After completing the evaluation, however, the NYPD chose instead to develop its own predictive policing algorithm.

The chart available here outlines what we know about the companies. It provides the variables each company used, public information about the companies’ clients, and relevant documents acquired as a result of the FOIL litigation. The variables were sent by each company to the NYPD as answers to a preliminary questionnaire before the trial evaluation. NYPD asked each company to list the data it needed from NYPD and any other data it planned to use. The client lists were compiled using company websites and public news stories.