A group of US lawmakers is pressing the United States Department of Justice to stop funding predictive policing systems through police grants unless the agency can show that the technology will not be used in discriminatory ways.
The concern centers on AI-based policing tools that claim to predict crime risks, but that lawmakers and researchers say can reproduce biased patterns already present in historical law enforcement data. According to the lawmakers, the DOJ has not provided evidence that it is checking whether grant recipients use federal money to buy these systems.
Why Congress is challenging DOJ grants
Seven members of Congress wrote in a letter to the DOJ, first obtained by WIRED, that the agency's answers so far have increased their concerns rather than reduced them. Their core point is direct: if the DOJ cannot tell whether state and local police agencies are buying predictive policing tools with federal funds, it cannot credibly say those tools are being used without discriminatory impact.
The funding at issue was awarded under the Edward Byrne Memorial Justice Assistance Grant Program. The Justice Department previously acknowledged that it had not tracked whether police departments used that funding to purchase so-called predictive policing tools.
The lawmakers' request is not framed as a general pause on all police technology. It is focused on systems that use data-driven predictions to guide policing decisions, particularly where those systems may reflect or intensify patterns of over-policing.
“We urge you to halt all Department of Justice grants for predictive policing systems until the DOJ can ensure that grant recipients will not use such systems in ways that have a discriminatory impact,” the letter reads.
Led by Senator Ron Wyden, a Democrat of Oregon, the lawmakers argue that the DOJ has an existing civil rights duty here. They say the department is required by law to “periodically review” whether grant recipients comply with Title VI of the nation's Civil Rights Act.
As they explain it, the DOJ is forbidden from funding programs that discriminate on the basis of race, ethnicity, or national origin. That concern applies whether the discriminatory outcome is intentional or not.
The accuracy and bias problem
The lawmakers' warning rests partly on a broader criticism of predictive policing software: historical crime data is not neutral simply because it is stored in a database. If the data reflects falsified crime reports, disproportionate arrests, or years of unequal police attention, a model trained on that data can carry those distortions forward.
Independent press investigations have found that popular predictive policing tools trained on historical crime data can replicate long-observed biases. The concern is that such systems may give police departments a technical-looking justification for practices that continue the over-policing of predominantly Black and Latino neighborhoods.
The source article points to an October headline from The Markup: “Predictive Policing Software Terrible At Predicting Crimes.” That story described researchers examining 23,631 police crime predictions and finding them accurate roughly 1 percent of the time.
For lawmakers, weak prediction performance and civil rights risk are connected. A tool that is inaccurate can still shape where officers are sent, who gets stopped, and which neighborhoods become the focus of enforcement. If those decisions then create more police records in the same places, the next round of data can make the pattern look more justified than it is.
“Pre dictive policing systems rely on historical data distorted by falsified crime reports and disproportionate arrests of people of color,” Wyden and the other lawmakers wrote, predicting—as many researchers have—that the technology serves only to create “dangerous” feedback loops.
The letter also states that “biased predictions are used to justify disproportionate stops and arrests in minority neighborhoods,” which then further biases statistics on where crimes occur.
What the lawmakers want examined
The lawmakers are asking that an upcoming presidential report on policing and artificial intelligence investigate predictive policing tools in the US. They want the review to look beyond whether a system exists or whether a department purchased it. Their request is about how these systems perform, how they can be understood, and what risks are hidden by limited visibility into how they are built.
They say the report should assess several core issues:
- accuracy and precision of predictive policing models across protected classes
- interpretability of the models
- validity of the models
- limits on risk assessment created by a lack of transparency from companies developing the tools
Those categories matter because predictive policing systems can affect real policing behavior even when their inner workings are hard to inspect. If a model cannot be interpreted, it becomes harder for agencies, courts, lawmakers, or the public to understand why certain areas or people are flagged. If a company does not provide enough transparency, it becomes harder to assess whether the tool is producing biased or unreliable outputs.
The lawmakers are not only asking for a study. They are also calling for conditions if the DOJ continues to fund this kind of technology after an assessment. In their view, the department should create “evidence standards” to identify which predictive models are discriminatory.
They say tools that fail those standards should be rejected for funding.
Who signed the letter
Senator Ron Wyden led the effort. The letter was also cosigned by Senators Jeffrey Merkley, Ed Markey, Alex Padilla, Peter Welch, and John Fetterman, along with Representative Yvette Clarke.
The shared concern across the letter is that federal grantmaking should not operate on blind trust when the technology involved may affect protected classes and policing outcomes. The lawmakers say the DOJ has not shown that it has reviewed whether grant recipients are complying with civil rights obligations in this area.
That gap is the central issue. If the department does not know whether grant money is being used to buy predictive policing software, it also does not know whether federally supported policing programs are relying on tools that may be inaccurate, biased, or opaque.
The lawmakers' demand puts the burden back on the DOJ: either demonstrate that predictive policing grants can be used without discriminatory impact, or stop funding those systems until meaningful safeguards exist.