Introduction 

In the summer of 2024, senior citizens residing in a Pennsylvania nursing home joined a growing list of thousands of voters whose registrations are being challenged by activists ahead of this year’s general elections.1 Several hundred miles away, in another mass challenge effort, David Sumrall, Head of the Bibb County Republican Party, urged the Bibb County Board of Elections to remove 159 Mercer University students from its rolls, a request which the board summarily dismissed.2 Just two years prior, Joseph Riggs, an unhoused voter living in Forsyth County, Georgia, was likewise swept up in a flawed mass voter challenge effort instigated by one man who filed a total of 31,500 voter challenges in that year alone, of which only approximately 200 resulted in immediate voter roll removals.3 For Mr. Riggs, the hurdles of fighting the voter registration challenge were too high, and he lost his opportunity to re-register before the hotly contested 2022 Georgia Senate elections.4 Unfortunately, frivolous mass voter registration challenges have not shown any signs of abating. In recent months, there has been a growing effort driven by activists—many continuing to espouse debunked voter fraud theories—to challenge the eligibility of thousands of registered voters.5

Emerging information about the data programs that mass challengers are using has prompted concern that the challenges rely on faulty data methodologies that put voters who lack traditional or long-term housing—such as nursing home residents, students who live in dormitories, unhoused people, and renters—at particular risk of having their voter registrations questioned and potentially canceled.6 This white paper will demonstrate that challengers who rely on programs such as EagleAI and IV3 are likely not only using flawed data methodologies but also running afoul of the robust legal framework protecting the fundamental right to vote. The paper will start by describing some of the emerging data programs which are being used to file mass challenges. It will then highlight some of the possible flaws in their general methodologies, which may impact a broad swath of voters and particularly those who lack traditional long-term housing. The paper will next describe the legal framework protecting qualified voters from list removal, including those who may lack stable or traditional housing. Lastly, it will provide a series of recommendations to assist election officials and voters who may confront mass voter challenges in the coming months.

Overview of Emerging List Maintenance Data Programs 

In the wake of the 2020 presidential election, activists espousing debunked voter fraud and election denial theories began developing data programs designed to detect “voter fraud” and remove individuals from state voter rolls.7 One such program was EagleAI Network (“EagleAI”).8 Created by Georgia doctor John W. “Rick” Richards Jr., EagleAI was marketed as a “tool of reckoning across the nation” to help validate, maintain, and review election rosters.9 Notably, a planning document obtained by investigative journalists claims that EagleAI uses multiple public government datasets to identify in-state, out-of-state, and foreign moves, including National Change of Address (NCOA) data.10 The NCOA database is a national file of individuals who have submitted a change of address request to the United States Postal Service (USPS).11 According to the planning document, EageAI cross-checks state voter roll data against NCOA data to determine whether a particular voter is registered to vote at a second address and thus should be challenged.12 In addition, the planning document suggests that EagleAI also uses address verification information from USPS’s Coding Accuracy Support System (CASS).13 Lastly, a report which uncovered an EagleAI training video suggests that the program interfaces outdated voter file data from VoteRef, a voter file database bankrolled by right-wing billionaire Dick Uihlein.14 According to journalists, once an EagleAI user identifies a voter record which they deem suspicious, they can, with a few clicks, generate a challenge ready for filing before a local election official.15

In addition to EagleAI, IV3 is another data program that has been promoted, mainly by True the Vote, as a tool that activists can use to challenge voter registrations.16 Investigative journalists who were able to analyze IV3’s public-facing code believe that the data program cross-references state voter registration databases with the NCOA database.17 According to their assessment, when IV3’s algorithm finds a “substantive discrepancy” between the two databases, it flags the record for review, potentially prompting the IV3 app user to submit a voter registration challenge.18 After analyzing the front-end components of IV3, journalists concluded that IV3 also gives users the ability to look up the registrations of all voters within the user’s county to assess for potential challenges.19

Methodologies of Data Analysis Programs IV3 and EagleAI 

Although programs such as EagleAI and IV3 are promoted as reputable tools for identifying voters who should be removed from state voter rolls, the available information about their methodologies suggests that, in truth, they are fundamentally flawed and unreliable. The section below will describe three ways in which EagleAI, IV3, and similar programs may be misusing USPS data in a manner that inadvertently flags qualified voters for removal, especially those who do not live in traditional, long-term housing. Of note, while state election officials use USPS data to help them maintain accurate and up-to-date voter rolls, they must follow a series of carefully outlined steps and procedures to prevent them from inadvertently removing qualified voters.20 On the other hand, the use of USPS data by private individuals—who lack expertise and training and are not required by law to conduct any due diligence—to determine voter eligibility is, at best, redundant and, at worst, capable of generating pressure to disenfranchise qualified voters in violation of federal law.

a. NCOA Data Possesses Inaccuracies 

First, any data program which relies on NCOA data to flag voters for mass challenges runs the risk of generating false positives, because the NCOA dataset in and of itself possesses inaccuracies. Indeed, NCOA data is comprised of approximately 160 million permanent change of address records constructed from the names and addresses of individuals, families, and businesses who have filed a change of address with the United States Postal Service.21 NCOA relies on the data submitted by millions of individual end users who enter their address change data.22 It is not quality-assured for errors, and thus, a small percentage of address errors in the NCOA system should be expected.23 In a disclaimer, USPS notably emphasizes the limitations of the NCOALink program: “The USPS makes no warranty or representation, either expressed or implied, with respect to the NCOALink® technology and/or the computer system in which it is contained, including its correctness, quality, performance, merchantability, or fitness for any particular purpose.”24 Unsurprisingly, recognizing the potential inaccuracies built into a massive dataset such as the NCOA, the NVRA permits states to use NCOA data to remove names from voter registration lists only if additional safeguards are used. Specifically, once a voter’s registration is matched to an NCOA request, that registration can be removed only if the voter has not responded to a mail notice sent by the registrar and has not voted or appeared to vote in two federal general elections.25 Otherwise, the NVRA only allows NCOA data to update a voter’s registration to their new address, not remove them.26 Programs which overly rely on NCOA data without any safeguards may thus inadvertently remove qualified voters.

b. NCOA Matches Do Not Necessarily Indicate That Voters Have Changed Residency

 In addition, programs which rely on USPS data may be at increased risk of producing false positives, because USPS data systems were designed for mail delivery, as opposed to voter list maintenance purposes. As described above, the NCOA is a database created by the USPS which tracks change of address information.27 The system allows customers to forward their mail when they move, either for a short period (for a college semester or a month at a summer college) or permanently.28 While IV3 and EagleAI are believed to use NCOA matches to flag voters who may have changed residency, these matches do not necessarily indicate that voters have moved permanently, if at all, beyond electoral precinct or district boundaries.29 The naïve interpretation of NCOA matches by lay users may thus inadvertently flag qualified voters, especially those who do not live in traditional long-term housing. Indeed, when an individual with access to NCOA, such as a postal worker, checks a name or address against the NCOA dataset, they receive a code that either indicates a “COA Match” or “No Match.”30 A “No Match” outcome indicates that the name and address provided to USPS did not return a new address record.31 A “COA Match” indicates that the record has a likely match in the database.32 USPS explains that there are eight different reasons why an address might match against the NCOA database: 

  1. A new address was found. 
  2. A new address was found outside of the USPS delivery area. 
  3. A change of address record was found without a new address provided. 
  4. A change of address record contained a P.O. Box address that was closed without a forwarding address provided. 
  5. A change of address record was found, but the new address on the COA record could not be converted to a deliverable address because it represents more than one delivery point. 
  6. A change of address record was found, but the new address could not be converted to a deliverable address. 
  7. A change of address record was found, but the new address cannot be ZIP + 4 coded. 
  8. A change of address record was found, but the change of address record had a secondary number, while the input address did not.33

Notably, while NCOA matches work adequately for mail delivery purposes, matching against the NCOA system for list maintenance purposes requires an additional level of sophistication that accounts for the specific NCOA match reason. A failure to conduct matches with the appropriate level of sophistication may result in substantial false positive matches, especially for individuals who lack long-term stable housing. Indeed, matching properly requires examination of the effective date of an address change, as well as the flag fields within the NCOA dataset indicating whether the change was permanent or temporary or was for an individual or a household. For example, without checking the flag fields indicating whether a change of address was permanent or temporary, an individual using NCOA data may inadvertently flag a voter who merely decided to temporarily forward their mail to another address. This is precisely what happened to Gamaliel Warren Turner Sr., a Georgia resident who requested that USPS forward his mail to a temporary address in California while he worked a temporary job.34 As a result of his request, he was flagged for removal by a mass voter challenge effort just before the 2021 Senate runoff election in Georgia.35 Like Turner, people who do not live in stable housing and need to forward their mail to friends, relatives, or a temporary housing facility may be at particular risk of being swept up in mass voter challenges because of similar misinterpretations of NCOA matches.

Similarly, without checking the flag field indicating whether a change of address is for a household or single person, a user of a data program such as EagleAI or IV3 may inadvertently flag an otherwise qualified voter. Take the hypothetical, for example, that one roommate in a college apartment with four other students fills out a change of address form to forward their mail when they graduate. In the process, however, the moving student accidentally mismarks the NCOA form as a “household” change, inadvertently forwarding his three roommates’ mail to his parents’ house in another state. A simple NCOA address match for those three additional roommates without any further investigation or analysis would flag them as potentially residing in another state. This type of misinterpretation using a dataset designed for mail delivery as opposed to list maintenance would also disproportionately impact individuals without stable long-term housing who may live with roommates. As a separate matter, NCOA’s file is partially cumulative. It contains 160 million rows. Not all 160 million rows are individual or recent moves. Individuals who move frequently without stable housing may appear on the file multiple times in the cumulative NCOA data file. If a lay user of a data program matches ONLY on address, they will likely capture further false positives. Lastly, compounding these potential pitfalls, reports suggest that EagleAI connects its users to outdated voter file data provided by VoteRef.36 VoteRef’s voter file purchase dates indicate that it uses voter data that, for many states, dates as far back as 2022.37 If an EagleAI user cross-references outdated voter files with NCOA data and lacks the requisite level of expertise to properly interpret the results, false positive flags are all but inevitable. Again, NCOA matches were designed to help the USPS deliver mail, not assist lay people to engage in list maintenance activities. Without a sophisticated analysis of the address, name, and flag fields of the NCOA database, NCOA matches may be easily misunderstood by users of programs such as IV3 and EagleAI and, as a result, flag qualified voters, especially those who do not live in stable, long-term housing.

i. CASS Certification Process was Not Designed to Identify Improper Voter Registrations.

 Similarly, the CASS certification process (Coding Accuracy Support System) is a computer algorithm designed by the USPS to standardize addresses.38 USPS created the CASS system to help standardize mailing addresses so as to reduce mailing costs. CASS works by breaking a standard address into elements. For example, this address: 

1456 East Elm Tree Lane, #455 Cambridge, Massachusetts 02138 

would parse as follows: 

Street Number: 1456 

Street Pre-directional: E (CASS would abbreviate “EAST” to “E”) 

Street Name: Elm Tree 

Street Type: LN (CASS has standardized abbreviations: “LN" for “Lane” and “ST” for “Street” “LN” etc.) 

Unit Type: APT (CASS standardizes “Apartment” and “#” to “APT”) 

Unit Number: 455 

City: Cambridge 

State: MA (CASS will shorten states to the 2-letter official postal abbreviation) 

Zip Code: 02138 

Zip4: 5544 (CASS will add the last four digits of the nine-digit ZIP based upon a lookup table. The last four digits of a Zip Code refer to a small geography, often one half of a street on one block. This allows mail to be pre-sorted in the order a mail carrier would physically deliver it.)

As mentioned above, a planning document obtained by investigative journalists suggests that EagleAI uses CASS as part of its list maintenance data analysis.39 Nevertheless, the use of CASS for list maintenance purposes may be misguided. Indeed, while CASS relies on reasonably clean data that allows USPS to deliver mail efficiently, voter files are often not formatted with the same level of precision. As a result, if users of programs such as EagleAI cross-check voter file addresses against CASS, they may mistakenly flag qualified voters whose voter registration addresses were inputted without the level of precision of the CASS certification process. For example, if a voter file has “Elm Tree” entered in as “ElmTree,” CASS may not find the address. That should not imply that the address does not exist or be grounds for a voter registration challenge. Likewise, CASS can be finicky about apartment number designations. If a voter file has “Room 459” recorded as “Room459,” it similarly may not appear in CASS due simply to the lack of a space. In this manner, the misuse of CASS may likewise subject qualified voters to challenges, especially those who may live in apartment building addresses. 

c. Use of Data Programs Such as IV3 and EagleAI by Lay Activists May Result in Large-Scale Misinterpretation of Minor Data Entry Errors 

Lastly, citizen activists with little list maintenance or data training are at risk of misinterpreting minor data entry irregularities caught by programs like IV3 or EagleAI as grounds for challenging a voter’s registration. In many jurisdictions, voter registration information is manually keyed into a database after voter registration applications arrive from the field.40 Because human data entry error is inevitable, voter registration records are bound to include some typographical errors like missing apartment or unit numbers, misspelled names, or mistyped birth years.41 Additionally, voter files typically contain two fields for a residential address: one for the street address (e.g., 25 Maple Street) and the second for the apartment or unit number (e.g., 36C).42 These address fields are often inconsistently populated in voter files.43 For example, a registrant’s street address and apartment or unit number may be entered into the first field (i.e., 25 Maple Street Apartment 3C) with no data appearing in the second.44

Investigative reports suggest that both IV3 and EagleAI match state voter registration data against NCOA data to identify discrepancies as possible grounds for a voter challenge.45 Unlike an election official who is well-versed in list maintenance practices and who is objectively analyzing a voter file, an untrained and misinformed lay activist of either program can easily mistake an inconsistency due to a mere clerical error as evidence of voter fraud. Voters who move more frequently, such as renters, students, low-income, urban families that are more likely to reside in apartment addresses, and those who live in temporary housing or homeless shelters where moves are frequent and room numbers vary over time may be especially impacted since there is more room for clerical errors or minor inconsistencies with frequent address changes.46 Notably, advocates have already drawn attention to how users of another data program called Check My Vote are misinterpreting clerical errors as evidence of voter fraud. In a similar manner to IV3 and EagleAI, it is believed that Check My Vote matches state voter registration files against a separate database of mailing addresses, specifically data it obtains from a direct-mail house known as “Melissa Data” which contains virtually no address errors.47 This has resulted in misinterpretations of clerical errors. When examining data from Lansing, Michigan, for example, Check My Vote identified 330 records for which an apartment was absent from the state voter file yet was listed in the Melissa database.48 While proponents of Check My Vote may point to these discrepancies as fraud, they are simply minor errors in large voter file datasets.49

Legal Framework for List Maintenance 

By using data programs which are likely predisposed to flagging qualified voters—especially those who happen to lack long-term housing—users of EagleAI and IV3 are likely not only engaging in flawed data analysis but also running afoul of the robust legal framework protecting the fundamental right to vote. To vote in a federal election within the United States, a person must be an American citizen, be at least 18 years old, and meet their state’s residency requirements. It is well established that states cannot limit the right to vote simply because a person lives in nontraditional housing—such as a homeless shelter or nursing facility—or moves frequently within state lines.50 Decisional law has found time and time again that the practice of limiting the right to vote to those who hold formal residential addresses violates the Equal Protection Clause and Due Process Clause of the 14th Amendment, the First Amendment, as well as the National Voter Registration Act.51 For example, in Pitts v. Black, the United States District Court for the Southern District of New York held that the New York City Board of Elections violated the Equal Protection Clause when it refused to allow homeless individuals to register to vote because they did not have fixed addresses.52 Similarly, in a class action involving a group of students residing in dormitories, the United States Court of Appeals for the Second Circuit held that a New York county election board violated the Equal Protection Clause when it denied students the right to register to vote on the basis that it had determined a campus dormitory was not a fixed, personal, or principal home.53

Federal law also makes clear that states cannot remove voters simply because they move within electoral precinct or district boundaries. Notably, to prevent state residents from being mistakenly removed from voter rolls merely because they have moved to different locations within precinct boundaries, the NVRA requires states to follow specific procedures before election officials may remove them on the basis that they believe the voter’s place of residence may have changed.54 Specifically, the NVRA prohibits states from removing registered voters on the basis that they may have changed their residence to another jurisdiction unless either 1) the voter confirms in writing that they have changed their residence to another jurisdiction, or 2) the voter has not responded to a mail notice sent by the registrar and has not voted or appeared to vote in two federal general elections.55 More broadly, the NVRA prohibits any systematic removal programs, including those initiated by mass challenge efforts that expunge voters from state rolls within 90 days of a federal primary or general election on the basis of a change in residency.56 By engaging in mass efforts to cull state voter rolls using data programs that are likely predisposed to flagging qualified voters who live in nontraditional or temporary housing, users of EagleAI who bow to their pressure also risk running afoul of the abovementioned protections.

Recommendations 

In the face of the unprecedented threat of individuals and groups filing mass voter challenges using programs such as EagleAI and IV3, we make the following recommendations: 

If you are an election official 

✔ Be aware that any lists of voter registrations flagged for removal by third parties may have been generated using flawed and unreliable data sources and software programs that are becoming widely available. Absent a requirement under state law, or an indication that the state’s current list maintenance practices are not sufficiently reasonable to facilitate secure elections, the presumption should be against utilizing such lists to remove voters. 

✔ If you decide to utilize third-party lists in any manner, it is imperative to first: 

  • Fully understand the data sources and programs used to identify voter registrations flagged for removal, and to understand how the third party selected which voter registrations to flag for removals. This is required to ensure that the state’s voter list maintenance is uniform, nondiscriminatory, and in compliance with the Voting Rights Act of 1965 as required by the NVRA. 
  • Ensure you are otherwise complying with federal law, including the NVRA’s prohibitions against systematic removals of voters within 90 days before a federal election and removal of voters on residency grounds without following the required notice procedures.57  
  • Use reliable data sources to conduct an independent investigation into any voter registrations flagged for list maintenance before making any changes to a voter registration and implement the safeguards (such as a waiting period of two federal elections, confirmation mailings, and additional voter notice) before flagging any voter registration for removal. 
  • If third-party lists are utilized for voter challenges, ensure you are complying with state law governing voter qualification challenges, as many states specify who can file a voter challenge, the procedures the challenger must follow, and the evidence, including personalized knowledge, upon which the challenge must be based.58

If you are a voter 

✔ Check your voter registration status early enough that you will have time to re-register before any state-mandated deadlines if you find you have been removed in error. 

✔ If you are a qualified voter and are incorrectly flagged in a challenge, do not panic—there are procedures that you can follow to dispute your challenge. 

✔ If you have questions or experience issues with your voter registration, contact your local election officials and call the nonpartisan Election Protection hotline at 866-OUR-VOTE for information and assistance. Bilingual support is also available in Spanish (888-VE-Y-VOTA), Arabic (844-YALLA-US), and Bengali, Cantonese, Hindi, Korean, Mandarin, Punjabi, Tagalog, Urdu, and Vietnamese (888-API-VOTE). 

Conclusion 

The threat of misguided mass voter qualification challenges this election cycle remains high. As this white paper has illustrated, activists using data programs like EageAI and IV3 are likely challenging the registrations of qualified voters simply because they move frequently or live in nontraditional housing. As a result, they risk running afoul of state and federal voting rights protections. Election officials should be wary of any lists produced by individuals or groups using software programs such as EagleAI and IV3, should always check the sources of these lists, and must ensure compliance with state and federal law governing challenges and other list maintenance activity. Voters should remain calm but vigilant, double-check their voter registration status, and reach out to election officials and Election Protection advocates for information and assistance.