Stay up to date with our Politics newsletter, sent weekly.
Two weeks after Trump officials said they had found more than 250,000 non-U.S. citizens on voter rolls in just four states, election officials are still struggling to understand where the numbers came from, and what the claims may portend for federal interference in this year's midterms.
Multiple state officials told NPR they fear these unverified numbers, or similar ones unveiled closer to the general election, could be used as pretext for warrants to allow law enforcement to interfere in vote-counting. Several reviews in recent years have found that actual voting by noncitizens is extremely rare.
The Department of Homeland Security, which oversaw the voter roll analysis of four states — California, Nevada, New Jersey and Pennsylvania — asked the states to respond to its allegations by Thursday. But the agency has not explained its methodology or provided the states with clarity on how it came to its conclusions.
"These numbers, until they're backed up by any kind of clear, transparent methodology, should be considered made up," said David Becker, the executive director of the Center for Election Innovation & Research.
At least three of the four states' top election officials wrote back to Homeland Security Secretary Markwayne Mullin, raising questions about the accuracy of the numbers and demanding answers on how DHS did the analysis.
Making matters even more puzzling, the Department of Justice previously sued all four states, along with more than two dozen more, for refusing to turn over versions of their voter rolls that include voters' partial Social Security numbers and driver's license numbers. Those lawsuits have so far been unsuccessful. Yet Mullin's letters to the states indicate his agency somehow obtained some of that sensitive data anyway.
The letters say DHS analyzed publicly available voter rolls in the four states. But public voter rolls don't include Social Security numbers, and the letters say the agency found around 118,000 voters combined in those states whose "name, date of birth, address and Social Security number match a non-citizen in our files" — raising questions about how DHS obtained voters' Social Security numbers.
"You're suing me because I wouldn't give you that data, that information. So where did you get it? How did you get it? And is it reliable?" asked Nevada Secretary of State Cisco Aguilar, a Democrat, in an interview with NPR.
The DOJ's lawsuit demanding Nevada's voter rolls with voters' personal data is still pending, while federal judges have so far dismissed the suits in California, New Jersey, Pennsylvania and 15 other states.
One possibility raised to NPR by election experts is that DHS relied on commercial data from a data broker to attempt to match Social Security numbers to the voter rolls, but the details are still unknown.
Previous reviews and data-matching efforts on voter rolls that lack a unique identifier like a Social Security number have been notoriously error-prone, Becker noted.
"There's no credible way to do this without creating an ungodly number of false positives," he said.
Furthermore, efforts to identify noncitizens on voter rolls frequently begin with inflated preliminary numbers that shrink down dramatically once investigations determine that many of the suspected noncitizens have become U.S. citizens.
Aguilar noted that DHS has not provided him with the names of the Nevada registered voters the agency claims may be noncitizens.
"Until I actually see the data, I can't do anything about it," Aguilar said.
"The thing we really fear"
DHS did not respond to NPR's questions about the methodology used to come up with the 250,000 figure.
Yet its allegations cast a shadow on more than just the four impacted states, as Election Day looms in fewer than 100 days.
In a July 17 press conference, Mullin promised more of these sorts of announcements and investigations in the weeks to come.
"We will pursue maximum pressure on this," he said. "We will be proactively looking at early voting and then after, post-election, we will continue to scrub all those that did vote."
Mullin also threatened prison time for state election officials who "choose not to participate in securing the elections."
One state election official, who spoke to NPR on the condition of anonymity because they were not permitted to speak publicly, said they worried the administration planned to use similar analyses after the election as a means to interfere with the vote-counting process.
"This is the thing we really fear," the official said. "Let's say they get a friendly judge somewhere to issue subpoenas the day after the election. So like Nov. 4, we have really close races [in our competitive districts] then the DOJ, FBI come knocking on one of our county doors and they say, 'We believe there's hijinks going on, we're going to confiscate all the ballots so we can count them.'"
Depending on the timing of such an order, the official said it's possible voting officials wouldn't be able to confidently know how many ballots were returned in the first place.
"They could do so much damage if they did that," the official said.
Similar fears have been circulating among voting officials since the FBI seized 2020 election ballots in Fulton County, Ga., in January. The search warrant was executed using an affidavit that relied on misinformation.
Justin Levitt, a law professor at Loyola Marymount University and a former Biden administration adviser, said he takes it "as a given that this administration is interested in sowing a little bit of chaos and in holding the door open … for potential interference."
He compared the 250,000 figure to a "prop" the administration could use to justify taking action and attempting to interfere.
"But that doesn't mean action is going to happen," Levitt said, noting that other steps would need to happen before ballots could be seized, such as convincing a federal judge to go along with an unprecedented plan. He said while it makes sense for election officials to plan for such possibilities, he believes they are unlikely.
DHS seems to acknowledge its numbers could be wrong
While a DHS fact sheet plainly stated that "over 250,000 noncitizens are illegally registered to vote" in the four states, Mullin's letters to the states were much less definitive.
Instead, they characterized the analysis as a "preliminary review" and the tally of noncitizens identified in each state as only a potential number. The letters also said a much smaller figure — 118,003 in the four states combined — had matched the name, date of birth, address and Social Security number of a noncitizen in DHS records.
"The most efficient way to ensure the accuracy of our findings is to work collaboratively on identity verification," the letters continued, seemingly acknowledging the numbers could be wrong.
The letters concluded by saying DHS would provide resources to help states "positively identify" noncitizens on their voter rolls.
The wording could refer to the use of a controversial DHS data system known as SAVE that the Trump administration overhauled last year and attempted to turn into a citizenship look-up tool.
In his July 17 news conference, Mullin warned states they would be punished for not using SAVE, but a federal judge in Washington, D.C., halted its use in late June, after finding that its creation was unlawful and that its results included inaccuracies that could disenfranchise eligible voters. The Trump administration is appealing that ruling.
NPR has obtained letters that California, Nevada and Pennsylvania secretaries of state sent back to Mullin. New Jersey did not respond to an NPR request to view its response. (Separately, New Jersey recently disclosed it discovered thousands of noncitizens on its voter rolls. State officials blamed a software glitch and said they are investigating.)
"I have serious doubts about the reliability of your data and accuracy of your claims, since my office has not shared our statewide voter registration list with DHS or any other federal agency or official," California Secretary of State Shirley Weber wrote, citing a state law that prohibits sharing voters' sensitive information.
The secretaries demanded detailed descriptions of the methodology DHS used as well as transparency about what state voter lists DHS analyzed.
California and Nevada asked if artificial intelligence tools were used in the analysis. Those letters also pointed out that the National Voter Registration Act bars election officials from systematically canceling voter registrations during the 90-day period before a federal election. The so-called "quiet period" for the November midterms begins next week, on Aug. 5.
Each of the states' responses expressed concern about improperly removing eligible voters from the rolls.
"Any collaborative effort with DHS must therefore be designed to minimize false positives, incorporate meaningful due process for affected registrants, and comply fully with Nevada law and the NVRA," Aguilar wrote in his response.
CEIR's Becker said the last-minute timing of Mullin's letters right before the quiet period, along with the failure to provide states with the names of the voters in question, are clues that this was not a good faith communication.
"If there was an administration that had legitimate information about individuals on a state's voter lists that were potentially ineligible, they would share the names and likely insist on an immediate response, particularly right before the 90-day quiet period begins next week," Becker said. "And they didn't do that."
Instead, Becker said he believed the point of the letters was political theater.
"There's nothing the states can do with the information in the letter, and [Trump administration officials] just plan to use this to amplify further disinformation about the election process," Becker said.
Possible clues about DHS' methods
While DHS has not detailed its methodology for coming up with the 250,000 figure, some clues into what may have happened can be found in a letter from the DOJ's top voting attorney earlier this year.
In the March 12 letter, Eric Neff, the acting chief of the DOJ's Voting Section, wrote to the former acting director of ICE, Todd Lyons, about a data sharing agreement between DOJ and DHS.
Neff explained that under the agreement, DOJ would provide DHS with public voter roll lists and DHS would "enrich" the data with DHS records so the data could be run through the SAVE system. Before a judge halted the use of the expanded version of SAVE, the system could be queried with a person's name, date of birth and either the last four digits of a Social Security number or a DHS immigration identifier number.
Neff's letter was made public in federal litigation that the voting rights group Common Cause and individual voters brought against the Department of Justice earlier this year.
One of the lawyers in that case, Nikhel Sus, chief counsel at Citizens for Responsibility and Ethics in Washington, told NPR the March 12 letter suggests that federal officials may have used a similar process with the public voter rolls from the four states. Under such a plan, federal officials could have tried to add Social Security data or other DHS identifiers to the public voter lists and then run the modified lists through SAVE before that system was halted.
Sus said such an approach, if it indeed happened that way, would not have led to accurate results, given the high potential for false matches.
"The 250K number is facially absurd. And it's no surprise the government has refused to disclose its underlying methodology — it would be impossible to reliably verify voters' citizenship status based solely on publicly available voter lists," said Sus, who also represents plaintiffs in the lawsuit challenging the government's overhaul of SAVE.
DOJ did not answer an NPR inquiry about whether the method described in Neff's letter was used to analyze the four states' voter rolls, but sent a statement that reads: "The Department will use every tool at its disposal to secure our elections and ensure noncitizens are not on the voter rolls. We are committed to working with our federal partners to identify and remove ineligible voters wherever they are found, and this work will continue in every state."
Federal procurement records that were published earlier this month and first reported by 404 Media also show that DHS is relying on records from the data broker Thomson Reuters, including in the agency's efforts to identify voter fraud. The company's CLEAR product and other datasets include extensive personal data on Americans.
NPR asked Thomson Reuters if the company knew whether its data tools were used in DHS' analysis of the four states' public voter rolls.
"As a matter of policy, we do not comment on specific customer contracts, and questions about an agency's mission and priorities are best directed to that agency," reads a statement from the company. It added that Thomson Reuters works with customers to support national security and public safety investigations.
"We prohibit the use of CLEAR for the purpose of identifying and locating noncriminal immigrants or undocumented individuals with the intention of deportation solely on the basis of the individual's immigration status. We take this restriction seriously, and we enforce it," the statement reads.
In his public comments, Mullin indicated that the data effort involving the four states was being led by Heather Honey, a prominent voice in election denial circles who was brought to DHS last year to oversee the federal government's election security work. Mullin said Honey was working to go through the thousands of voter records individually now, to figure out who among the alleged noncitizens actually voted.
"We will go through those records one by one, and we will pursue everybody," Mullin said.
Copyright 2026 NPR
