Yes, a lawsuit against Pasco County Sheriff’s Office confirmed that the PredPol predictive policing software reinforced racial bias in patrol patterns. In December 2024, just before trial was set to begin, the Pasco County Sheriff’s Office settled with four residents for $105,000, admitting to Fourth Amendment and First Amendment violations tied to its use of the “Intelligence Led Policing” (ILP) program powered by PredPol. The settlement revealed that approximately 1,000 people—many of them minors—had been monitored under a system designed to predict future crime but actually predicted where police would patrol, cementing existing biases embedded in arrest data.
This settlement is part of a larger reckoning with predictive policing across the country. The case demonstrates exactly how the algorithm’s core flaw operates: it ingests historical arrest patterns (which are already skewed by decades of biased policing) and uses those to identify “high-risk” areas for increased surveillance—meaning more arrests, which feed back into the next cycle. It’s a feedback loop that doesn’t reduce crime; it manufactures the appearance of crime in the communities already most heavily policed.
Table of Contents
- How Predictive Policing Software Like PredPol Is Supposed to Work
- The Pasco County Case and the Constitutional Violations Admitted
- What Research Reveals About PredPol’s Core Algorithm Flaw
- How Other Police Departments Responded to Predictive Policing Evidence
- PredPol’s Rebranding and What Happened to the Company
- Legal Implications and Emerging Lawsuits Against Predictive Policing Programs
- What This Means Going Forward for Predictive Policing
How Predictive Policing Software Like PredPol Is Supposed to Work
Predictive policing tools like PredPol (which rebranded to Geolitica in 2021) are software systems designed to forecast where crime is most likely to occur. The pitch is straightforward: if police can predict hot spots with data analysis, they can allocate resources more efficiently and prevent crimes before they happen. Departments feed the software historical crime data—arrest records, incident reports, 911 calls—and the algorithm identifies geographic areas where the software recommends increased patrols. The fundamental problem is what goes into the model.
Crime data is not an objective record of where crime actually occurs; it’s a record of where police have already decided to look. Communities that have been over-policed for decades have arrest records that reflect enforcement intensity, not crime prevalence. When an algorithm learns from that data, it doesn’t learn “where real crime happens”—it learns “where arrests have been made.” A low-income neighborhood with heavy police presence will show high arrest rates; an affluent neighborhood with light police presence will show low arrest rates, even if both have similar underlying crime. The software inherits and amplifies those existing biases.
The Pasco County Case and the Constitutional Violations Admitted
The Pasco County Sheriff’s Office began its Intelligence Led Policing program using PredPol to identify areas where it should concentrate patrols. In practice, this meant deputies would perform frequent “prolific offender” checks at the residences of individuals the algorithm flagged. They would show up at homes without warrants, knock on doors, and question residents—sometimes multiple times per month. Roughly 1,000 people were enrolled in the program, and many of them were juveniles.
The settlement, finalized in December 2024, included the sheriff’s office admitting to two specific constitutional violations. First, the prolific offender home checks violated the Fourth Amendment by exceeding the scope of constitutionally implied license rights—police cannot simply knock on someone’s door as often as they want without some lawful purpose or consent. Second, the program violated the First Amendment by punishing people based on their “intimate associations.” Because deputies visited the homes of flagged individuals regardless of whether those individuals had personally committed recent crimes, they were penalizing people for living in the same house as someone the algorithm targeted. The settlement included $105,000 paid to four named plaintiffs, and the acknowledgment that roughly 4,000 door-knock interactions had occurred under this program.
What Research Reveals About PredPol’s Core Algorithm Flaw
The Pasco County case aligns with what academic and investigative research has consistently shown: PredPol is predicting future policing, not future crime. A massive data leak analyzed by researchers confirmed that the algorithm does not identify crime hotspots better than random chance—it simply replicates and intensifies existing arrest patterns. The Markup investigated the system in 2023 and found that PredPol’s success rate in predicting crime was less than 0.5%, meaning the software was essentially performing at random. The implication is stark: the tool doesn’t make policing safer or more efficient.
Instead, it provides a technological veneer of objectivity to what is fundamentally a system for automating biased enforcement decisions. When a jurisdiction has a history of policing a particular neighborhood more intensely (whether due to explicit bias, resource allocation decisions, or historical patterns), PredPol learns that history and recommends even more policing there. Residents get visited more often, arrested more often, and therefore appear in the arrest data more often, which trains the next iteration of the model to target that same neighborhood harder. The cycle reinforces itself automatically, and the department can point to the software as an objective decision-making tool rather than acknowledge the human and historical biases baked into the data.
How Other Police Departments Responded to Predictive Policing Evidence
The Los Angeles Police Department discontinued its use of PredPol in April 2020, after mounting evidence and community pressure highlighted its role in driving racist policing patterns. LAPD’s decision came years into the tool’s deployment and followed years of criticism from civil rights organizations, but it represented a significant shift: one of the nation’s largest police departments publicly acknowledged that the tool did more harm than good. That withdrawal was only the beginning of a broader reassessment. In January 2024, seven Democratic senators sent a letter to the U.S.
Department of Justice demanding that the agency halt funding to predictive policing programs. The senators cited “mounting evidence indicates that predictive policing technologies do not reduce crime” and raised concerns about civil rights violations. This federal-level pressure signals that police departments considering these tools now face political and reputational risks. However, some jurisdictions have continued to use PredPol or switched to newer versions or competitors offering similar promises. The software market has adapted, with vendors rebranding and marketing updates designed to address bias concerns—though the fundamental flaw (learning from biased historical data) remains.
PredPol’s Rebranding and What Happened to the Company
In 2021, PredPol rebranded itself as Geolitica, a shift that coincided with growing public scrutiny of predictive policing. The rebrand appeared designed to distance the company from the criticism that had accumulated under the PredPol name. However, investigative reporting by The Markup in 2023—looking at the newer Geolitica system—found that the fundamental problem persisted: the software had a documented success rate of less than 0.5% at predicting actual crime locations.
The rebranding illustrates how predictive policing companies have adapted to criticism: they change names, claim to use “updated” algorithms, and promise that newer versions address bias concerns. However, as long as the input data consists of historical arrest patterns from biased systems, the core problem remains. A more sophisticated algorithm cannot fix a poisoned data source. Some versions of the software now include what vendors call “bias audits” or “fairness metrics,” but independent researchers have found these tools often provide false confidence without actually preventing discriminatory outcomes.
Legal Implications and Emerging Lawsuits Against Predictive Policing Programs
The Pasco County settlement establishes important legal precedent. By admitting to Fourth Amendment and First Amendment violations, the sheriff’s office created a roadmap for plaintiffs’ attorneys suing other jurisdictions using similar tools.
Civil rights organizations including the Institute for Justice have used the Pasco case to support other litigation challenging predictive policing programs in different jurisdictions. The question is no longer “is predictive policing theoretically risky?” but rather “has your jurisdiction admitted or will a court find that your use of these tools violated residents’ constitutional rights?” Residents of communities targeted by predictive policing programs now have both individual claims (if they were personally harmed by illegal stops or searches) and potential class action claims (since the programs typically targeted hundreds or thousands of people). The settlement amounts in Pasco County—$105,000 split among four people—suggest that if larger class actions succeed, payouts could be substantial.
What This Means Going Forward for Predictive Policing
The convergence of the Pasco County settlement, the congressional pressure, the LAPD discontinuation, and The Markup’s investigative reporting indicates a turning point in how police departments and policymakers view these tools. The narrative that predictive policing is a neutral, data-driven alternative to biased human decision-making has collapsed. What remains is the recognition that these systems automate and amplify existing biases while claiming objectivity.
Looking ahead, the question is whether other jurisdictions will proactively abandon predictive policing tools or wait for lawsuits and settlements to force the issue. Some departments are exploring different approaches to resource allocation that don’t rely on historical arrest data. Others are simply reverting to traditional patrol strategies while maintaining the appearance of innovation. For residents of communities that have already been subjected to these programs, the Pasco County case offers a template for seeking compensation and justice.
You Might Also Like
- Lawsuit Claims Whoop Fitness Tracker Shared Heart Rate Variability Data With Life Insurers
- Lawsuit Claims Thomson Reuters Sold Journalists’ Source Emails to Government Without Warrant
- Lawsuit Claims Sonos Smart Speaker Collected Audio Snippets Without Activation Word Being Spoken