HAVA Administrative Complaint -- Bernegger v. Carew et al.
BEFORE THE TEXAS SECRETARY OF STATE
ELECTIONS DIVISION
AUSTIN, TEXAS
ADMINISTRATIVE COMPLAINT
Pursuant to Section 402 of the Help America Vote Act of 2002 (P.L. 107-252; 52 U.S.C. §15512) and 1 Texas Administrative Code §81.171
In the Matter of:
PETER BERNEGGER, Complainant, v. MICHELE CAREW, Elections Administrator, BEXAR COUNTY ELECTIONS DEPARTMENT, JANE NELSON, Texas Secretary of State,
Respondents.
I. JURISDICTION AND STANDING
Complaint No.: ___________
Date Filed: March 1, 2026
VIOLATIONS OF TITLE III OF THE HELP AMERICA VOTE ACT OF 2002
HEARING ON THE RECORD REQUESTED
1. This complaint is filed pursuant to Section 402 of the Help America Vote Act of 2002 ("HAVA"), P.L. 107-252, codified at 52 U.S.C. §15512, which requires each State receiving HAVA funds to establish and maintain uniform, nondiscriminatory administrative complaint procedures through which any person who believes there
is a violation of any provision of Title III of HAVA may file a complaint. Texas accepted federal HAVA funds and adopted such procedures. This complaint invokes those procedures.
2. Title III of HAVA applies to federal elections and mandates statewide voter registration list maintenance standards under 52 U.S.C. §21083 (Section 303), provisional voting safeguards under 52 U.S.C. §21082 (Section 302), and voting system standards under 52 U.S.C. §21081 (Section 301). The election at issue is the March 3, 2026 Republican Primary Election in Bexar County, Texas, which includes the federal office of U.S. House, Texas 21st Congressional District, encompassing Bandera, Bexar, Blanco, Comal, Gillespie, Hays, Kendall, Kerr, and Real counties.
3. Texas adopted its HAVA State Plan in January 2005 (the "State Plan"). The State Plan remains the operative document governing administrative complaint procedures. Section 9 of the State Plan states that the Secretary of State "has adopted an administrative complaint procedure through its rulemaking authority" and that "[c]omplaints are limited to those arising from violations of Title III of HAVA." The implementing Texas administrative rule at 1 Texas Administrative Code §81.171, adopted effective November 13, 2003 and amended effective July 1, 2004, provides the procedural framework.
4. The Texas Secretary of State is the proper adjudicating authority. The State is required to make a final determination within 90 days per HAVA §402(a)(2)(H) and 1 TAC §81.171(i).
5. Complainant Peter Bernegger is a citizen of the United States, a registered voter/elector, and a qualified elector whose rights are directly affected by the integrity, accuracy, and auditability of the statewide voter registration list used in federal elections. Complainant files this complaint on the basis of publicly available forensic evidence, publicly available official records, and the independent expert analyses described herein. He has direct knowledge of the data, communications, with his and colleagues analysis of the information and data of the Texas Secretary of State, of Bexar County election data sent out on or about February 18, 2026.
II. RESPONDENT
6. Pursuant to 1 TAC §81.171(a)(2), the following government officials and instrumentalities are named as Respondents. Each bears responsibility for one or more HAVA Title III violations alleged herein through direct operational authority, certification authority, supervisory responsibility, or any combination thereof.
7. Respondent 1: Michele Carew, Elections Administrator, Bexar County, Texas. Official address: 1103 S. Frio, Suite 100, San Antonio, Texas 78207. Telephone: (210) 335-8683. Respondent Carew is the county officer responsible for voter registration activities and election operations in Bexar County, serving as the Early Voting Clerk. She holds direct operational authority over the KnowInk electronic pollbook system -- including the Poll Pad devices, ePlus control stations, and cloud database connectivity -- and all early voting data files at issue. She recommended the KnowInk procurement to the Commissioners Court on May 13, 2025.
8. Respondent 2: Bexar County Elections Department. Official address: 1103 S. Frio, Suite 100, San Antonio, Texas 78207. The instrumentality responsible for administering the March 3, 2026 Republican Primary Election, including procurement, deployment, configuration, and maintenance of the KnowInk system and management of all early voting check-in records.
9. Respondent 3: Jane Nelson, Texas Secretary of State. Official address: James E. Rudder Building, 1019 Brazos Street, Austin, Texas 78701. Mailing address: P.O. Box 12697, Austin, Texas 78711-2697. Secretary Nelson is the chief election officer for the State of Texas, responsible under Texas Election Code §31.014 for prescribing standards for and certifying electronic pollbook systems. Her office certified the KnowInk PollPad system on four occasions (versions 2.4.9, 3.6, 3.6.3, and 4.1.3), most recently October 8, 2025, each time relying on Pro V&V, Inc. test reports. The Secretary has supervisory authority over all county election officials and bears responsibility for ensuring certified technology meets HAVA Title III requirements. The Secretary is also the adjudicating authority for this complaint under 1 TAC §81.171. Respondent Nelson's office is aware that the KnowInk PollPad does not provide hash validation of its software or data files, a requirement identified in the EAC's own publications and mandated for voting systems under Texas Election Code §129.023(c-1) (SB 1, 87th Legislature), yet has not required this safeguard for electronic pollbooks.
III. BASIS FOR NAMING GOVERNMENT RESPONDENTS
10. HAVA Title III imposes affirmative obligations on the State and its political subdivisions. Section 402(a)(2)(B) of HAVA provides that "any person who believes that there is a violation of any provision of title III (including a violation which has occurred, is occurring, or is about to occur) may file a complaint." The statute authorizes complaints alleging Title III violations by election officials and the entities they administer. The Texas implementing rule at 1 TAC §81.171 requires a description of the alleged violation "sufficient to apprise the Secretary of State of the nature and specifics of the complaint" (§81.171(c)(3)).
11. Respondents 1 and 2 bear direct operational responsibility for the HAVAcompliant conduct of the March 2026 Primary, including security of the computerized voter registration list interface (§303(a)(3)), voter registration record integrity (§303(a)(4)), and audit capacity (§301(a)(2)). Their office procured, deployed, and operated the KnowInk system from which the anomalous data originated.
12. Respondent 3 is the chief state election official responsible for certifying electronic pollbook systems under Texas Election Code §31.014 and overseeing HAVA compliance statewide. The Secretary's certification of the KnowInk system is a state action that enabled the deployment of equipment in which the violations occurred. The Secretary's failure to require hash validation for electronic pollbooks -- despite requiring it for voting systems under Texas Election Code §129.023(c-1) -- constitutes a regulatory gap that directly facilitated the undetected anomalies.
13. This complaint references the conduct of KnowInk, LLC (the system vendor), Scott Leiendecker (its founder and CEO), and Pro V&V, Inc. (the testing laboratory) as factual evidence supporting the violations alleged against the government Respondents. These private entities are not named as Respondents because HAVA Title III imposes obligations on the State and its officials, not on private vendors. However, the factual conduct of these entities is integral to the government Respondents' violations and is set forth in the Statement of Facts for evidentiary purposes.
IV. REQUEST FOR HEARING ON THE RECORD
14. Pursuant to HAVA §402(a)(2)(E) and 1 TAC §81.171(c)(4), Complainant hereby requests a hearing on the record to present documentary evidence, witness testimony, and oral argument in support of each allegation contained in this complaint.
V. STATEMENT OF FACTS
The following facts are alleged in support of this complaint. The factual findings are drawn from the following sources, each incorporated by reference:
(1) Bexar County Forensic Analysis, version 20, by Dr. Andrew Paquette, dated February 27, 2026 ("Forensic Report v20");
(2) Bexar County Forensic Analysis, version 18, by Dr. Andrew Paquette ("Forensic Report v18");
(3) Public address of Congressional candidate Weston Martinez, February 24, 2026, across from the Alamo, San Antonio, Texas;
(4) Martinez Campaign Press Release, February 26, 2026;
(5) Independent expert analyses by Dr. Walter Daugherity (Texas A&M University) and Dr. Andrew Paquette;
(6) Texas Secretary of State Electronic Pollbook Certification records for KnowInk, LLC (versions 2.4.9, 3.6, 3.6.3, 4.1.3);
(7) EAC Voluntary Electronic Poll Book Certification Requirements (VEPBCR 1.0), April 8, 2024;
(8) EAC publication: "What is Hash Validation and Why Should Election Officials Care?" July 12, 2021;
(9) EAC Best Practices for Election Technology;
(10) Texas Election Code §129.023(c-1) (SB 1, 87th Legislature) regarding hash validation;
(11) Texas SOS Advisory 2022-30: Hash Validation Procedures for Logic and Accuracy Testing;
(12) Collaborative field investigation by the Martinez campaign and independent election integrity researchers including Lori Gallagher, Dr. Laura Pressley, Dr. Walter Daugherity, and Dr. Andrew Paquette;
(13) Texas HAVA State Plan, January 2005;
(14) Help America Vote Act of 2002, P.L. 107-252.
A. The Source Data and Its Discovery
15. On February 19, 2026, Bexar County early voting poll pad export data for the TX-21 Republican Primary was captured and distributed to precinct chairs by Congressional candidate Weston Martinez and election integrity researcher Lori Gallagher (@Lorionafarm). The data was obtained through public records. The capture occurred before the February 25, 2026 bulk file replacement event -- a fact that proved critical, as the replacement files contain no trace of the anomalous records described herein.
16. The source file is the Bexar County Republican primary early voting check-in file, containing records of voters who had checked in as of February 18, 2026. The file includes the following fields for each record: Voter Name, State Voter Identification Number ("State ID"), Street Address, Precinct Number, Party Affiliation, and Check-In Date.
B. Dataset Composition
17. The original dataset contained 8,923 voter check-in records spanning two dates: 4,078 records dated February 17, 2026 (day one of early voting) and 4,845 records dated February 18, 2026 (day two).
18. Of the total 8,923 records, 4,813 records contain legitimate integer State Voter IDs (formatted as ten-digit whole numbers). The remaining records contain State IDs bearing fractional (decimal) components to five decimal places. The composition of records by date is set forth in the following table:
19. Texas State Voter Identification numbers are administrative integers -- whole numbers assigned from a statewide pool. They have no legitimate fractional component. A fractional State ID -- such as 1256119003.57755 or 1272351347.66981 -- is not a rounding artifact, display error, or migration residue. It is a mathematically impossible value in any legitimate voter registration system.
20. Every one of the fractional State IDs was verified by automated comparison against the Texas statewide voter database (TEAM) as of a September 8, 2024 snapshot. No fractional State ID appears in any statewide TEAM snapshot reviewed. Dr. Paquette concludes these synthetic registrations never went up to or into the county's voter registration list or the State's -- they existed only within the KnowInk system.
C. Clarification of Record Count: 4,110 Versus 4,117
21. The full check-in file contains 4,117 records bearing fractional State IDs. A filtered extract used for the definitive gap analysis contains 4,110. This discrepancy of seven records is attributable to header artifacts and malformed rows that were excluded during data validation. The validated count of unique fractional identifiers used for all mathematical proofs in this complaint is 4,110. The sevenrecord discrepancy does not affect any finding or conclusion.
D. The KnowInk System Architecture
22. The data at issue originated from the KnowInk electronic pollbook system deployed in Bexar County. The KnowInk system is not a single device but an integrated architecture comprising multiple components: (1) Poll Pad tablet devices used by poll workers at voting locations to check in voters; (2) the ePlus control station used for system administration, election configuration, and data management; (3) a cloud-based database infrastructure providing remote data synchronization, real-time voter status updates, and centralized data storage; and (4) communications network connectivity linking these components.
23. It has not been determined which component of the KnowInk system was the vector through which the anomalous records were introduced -- the anomaly could have occurred via the Poll Pad devices, the ePlus control station, the cloud database, or another access point within the architecture. All components are designed, manufactured, and maintained by KnowInk, LLC.
24. The KnowInk Poll Pad devices actively connect to communications networks during the conduct of elections, transmitting check-in data to and receiving data from external servers -- including state voter rolls -- every 15 minutes during active voting. This network connectivity during the conduct of a federal election creates a real-time attack surface.
E. The Constant Interval: Mathematical Proof of Deliberate Construction
25. When all 4,110 fractional-ID records are sorted in ascending order by State ID and the gap between each consecutive pair is calculated, an extraordinary pattern emerges. At rounded precision, every gap is identical: 22,084.82189.
26. The first record in the injected sequence is Robert Earl Abel (State ID 1,253,115,467.79993). The last record is Joseph Edward Braswell (State ID 1,343,862,000.96332). The total span is: 1,343,862,000.96332 minus 1,253,115,467.79993 = 90,746,533.16339
27. Dividing the total span by the inter-record gap: 90,746,533.16339 divided by 22,084.82189 = 4,109.0000 (exactly)
28. This result -- a perfect integer with zero remainder -- is the definitive proof of deliberate construction. A randomly generated or accidentally corrupted sequence cannot produce this result. When a total span divided by a uniform gap yields an exact integer equal to the count of intervals (4,110 records produce 4,109 intervals), the sequence is the output of deliberate computation.
29. A demonstrative table showing 14 consecutive records and their identical gaps is attached as Exhibit A.
F. The Four Gap Values: Floating-Point Arithmetic Fingerprint
30. At full 10-digit decimal precision, the 4,109 individual gap values resolve into exactly four distinct values, all differing only at the 10th decimal place:
D 22,084.8219001293
Forward
Reverse
Reverse
31. The four values form two micro-pairs: {A, B} constitute the forward-pass pair and {C, D} constitute the reverse-pass pair. This pairing is consistent with IEEE 754 double-precision floating-point arithmetic, in which the same division operation executed in forward versus reverse order produces conjugate rounding residuals due to the non-commutativity of floating-point addition at the limit of representable precision.
32. The four values are organized into palindromic seven-element blocks of the form A B A B C D C (forward pass) followed by C D C B A B A (reverse pass). This ABABCDC / CDCBABA structure repeats throughout the full 4,109-gap sequence. The palindrome arises because the algorithm processes the 735-record anchor list in alternating forward and reverse passes.
33. Ten boundary anomalies -- locations where the palindromic pattern breaks -occur at regular intervals corresponding exactly to the 735-record cycle boundaries. These anomalies independently confirm the anchor group size of 735 without reference to any other data field.
34. The palindromic gap structure cannot arise from manual data entry, database migration, display formatting, software misconfiguration, or random corruption. It is the deterministic output of a specific floating-point loop executing in a compiled language, processing a fixed-size record set in alternating directional passes. The algorithm was not a spreadsheet formula. It was purpose-written software.
G. The 735 Anchor Voters and Clone Group Structure
35. The 4,110 fractional records correspond to 735 genuine registered voters -- real individuals who actually checked in and voted on February 18, 2026. Each of these
735 anchor voters was replicated either 5 or 6 times to produce the 4,110 derivative records: Synthetic Copies
36. The distribution of 5-copy and 6-copy groups is not arbitrary. It is the unique solution to a two-equation integer constraint system: 300 groups of 5 copies plus 435 groups of 6 copies is the only whole-number solution that yields exactly 4,110 total records from exactly 735 anchors.
37. When the 4,110 fractional records are sorted by State ID, they resolve into exactly six complete or partial passes through the 735 anchor names in strict alphabetical order by last name. Copies 1 through 5 are each complete runs of all 735 anchors. Copy 6 begins the same way but terminates at record 435 with Joseph Edward Braswell. The cutoff falls precisely between two members of the same household. A legitimate check-in file does not contain six sequential passes through an alphabetically sorted voter list. A database generation script does.
38. The intra-anchor gap -- the distance between clones of the same voter -- is 16,232,344.09226, first identified by Dr. Walter Daugherity (Texas A&M University). This value equals exactly 735 multiplied by 22,084.82189, confirming the algorithm distributed 4,110 records across 735 anchors using a single spacing constant. A representative sample of anchor voters and their clone patterns is shown in Exhibit C.
H. Alphabetical Distribution: Statistical Impossibility
39. When the 735 February 18 integer (anchor) records are sorted by State ID and the last name of each voter is extracted, 731 of 735 records -- 99.5% -- have last names beginning with A, B, or C. The remaining four are compound name parsing artifacts. The true proportion is effectively 100%.
40. The February 17 control group shows what organic voting data looks like: 4,078 check-ins with last names distributed across all 25 letters in natural proportions, with A/B/C surnames constituting approximately 18.7% of the population. The February 18 integer records show none of this natural distribution:
Metric Feb. 17 (Control) Feb. 18 Integer IDs
Expected if organic? Yes
Statistically impossible
41. The anchor names end at "Chicas" -- the last C-name in the sorted voter roll before the D-names begin. This boundary is where the counter reached 735. Automated comparison confirms that every C-name through Chicas is present in the anchor group, and every C-name from Chicoine onward is absent -- an exact alphabetical cutoff with no exceptions. If A/B/C names represent approximately 20% of the full voter roll, the full organic check-in population on February 18 would have been approximately 3,675 to 4,400 voters. The injection replaced the D-through-Z portion of that real list with 4,110 fractional clones.
I. Address Fabrication Patterns
42. Two distinct address assignment strategies are present within the synthetic records. Sixty-seven percent of groups (494 groups) use sequential +1 house numbers: the house number increments by exactly one for each successive synthetic copy. Thirty-three percent of groups (239 groups) use identical addresses: all synthetic copies for a given voter share the exact same address. Only 2 groups out of 735 (0.3%) show any deviation from these two patterns, consistent with a sign-inversion bug in the generating algorithm triggered by processing two adjacent same-street addresses simultaneously.
43. The choice of address pattern is statistically independent of the choice of clone count (5 versus 6), indicating separately controlled variables within the generating algorithm.
J. The Dead Zone: Fraudulent IDs in Empty ID Space
44. All 4,110 fractional IDs occupy a contiguous block of the State ID space ranging from approximately 1,253,115,468 to 1,343,862,001 -- a span of roughly 90.7 million consecutive ID numbers. Not one of the 735 February 18 integer State IDs falls within this range. The nearest integer record is 34 million ID units away from the fractional minimum.
45. The fraudulent records were inserted into a dead zone -- a band of approximately 90 million consecutive ID numbers containing no legitimate voters whatsoever. Selecting this specific sub-range required prior knowledge of which portions of the statewide ID space were unoccupied. That is reconnaissance, not improvisation.
K. User Interface Concealment
46. The KnowInk Poll Pad display truncates State Voter ID numbers to their integer portion. Poll workers viewing the check-in screen would see a synthetic record with State ID 1256119003.57755 displayed as "1256119003" -indistinguishable from a legitimate integer ID. The fractional component that identifies every anomalous record is invisible in normal Poll Pad operation. Detection required export of the raw underlying data files and independent forensic analysis.
L. Ballot Generation Implications
47. Each of the 4,110 anomalous records would have generated a ballot authorization event in the poll pad system. Since ballots carry no voter identification information once separated from their envelope or stub, any ballots cast under these records are indistinguishable from legitimate ones and cannot be identified or removed from the count.
48. The KnowInk system transmits check-in status every 15 minutes to prevent double voting. Five or six identical check-ins per voter identity is incompatible with normal system operation. Either the records were inserted as pre-checked-in entries bypassing the state verification loop, or the anti-double-voting safeguards were not functioning.
M. The Built-In Deletion Mechanism
49. The use of fractional State IDs was not merely a byproduct of the anomaly -- it functions as a built-in cleanup mechanism. Following the voting period, the fractional component provides a precise, machine-executable search key: a single database query filtering for non-integer State IDs would return exactly these 4,110 records and no others. Execution of that query would batch-delete all anomalous records, leaving no trace in the permanent voter roll. The permanent roll would then show 735 legitimate voters who each voted once, while the ballot box would contain up to 4,110 additional ballots cast in their names. Because ballots are anonymous after casting, the extra ballots cannot be identified or removed from the count.
N. The Bulk File Replacement Event
50. On February 25, 2026, all 14 CSV early voting files (7 Republican, 7 Democratic, spanning February 17-23) underwent simultaneous bulk replacement in a synchronized 78-second automated operation. The creation timestamps on all 14 files read 08:46:26; the modification timestamps read 08:47:44. The postreplacement files contain zero fractional IDs.
51. The original data survives only because it was captured on February 19 and distributed to precinct chairs before the replacement. Had precinct officials not distributed a copy of the February 19 file before its erasure, all traces would have vanished.
52. A detailed timeline of the 14 replacement files is attached as Exhibit D. No error logs, incident reports, or change-control records have been disclosed for the bulk file replacement. The bulk replacement permanently broke the chain of custody for the entire Bexar County early voting dataset. No file in the current production dataset can be considered forensically intact.
O. Verification of Affected Voters
53. Congressional candidate Weston Martinez personally contacted voters whose names appeared in the duplicated records. Each confirmed voting only once. The multiple check-in entries were not authorized by these voters. Martinez stated
publicly: "I went out and talked to some of these people on the list. I confirmed -they voted. But they only voted once. So why does the list show them on there multiple times? That's not error. That's fraud."
P. Expert Verification
54. Dr. Andrew Paquette (primary forensic analyst) and Dr. Walter Daugherity (Texas A&M, independent verification) both independently conclude the data is the product of deliberate, purpose-written software. Dr. Paquette states: "The simultaneous satisfaction of every one of these constraints by an accidental process is not merely improbable. It is structurally incoherent. Probability applies when an outcome is conceivable but unlikely. Here, there is no mechanism -- no class of software failure, database corruption, or system misconfiguration -- that could generate this structure."
Q. KnowInk Certification History
55. The Texas Secretary of State certified the KnowInk PollPad on four occasions, each relying on test reports from Pro V&V, Inc., a NIST-certified testing laboratory:
Jan. 23, 2020
2.4.9 Keith Ingram (under Secy Hughs)
Aug. 21, 2024 PollPad 3.6
Christina Worrell Adkins (under Secy Nelson)
Mar. 6, 2025 PollPad 3.6.3 Christina Worrell Adkins (minor update)
Oct. 8, 2025
PollPad 4.1.3
V&V
Christina Worrell Adkins (under Secy Nelson)
56. The certification process did not test for: (a) resistance to back-end data injection; (b) detection of fractional or impossible State ID values; (c) adequacy of the anti-double-voting mechanism against bypass; (d) immutable audit trail
integrity; or (e) hash validation of data files. These are the precise vulnerabilities at issue in this complaint.
57. Pro V&V certified the ES&S ExpressPoll on August 21, 2024 (the same day as KnowInk PollPad 3.6). That system was decertified by the Texas Secretary of State in December 2024 -- only four months later -- after failing seven technical and three functional standards during the November 2024 General Election in Dallas County.
R. Hash Validation Failure
58. The EAC published "What is Hash Validation and Why Should Election Officials Care?" (July 12, 2021), stating that hash validation provides a "digital fingerprint" ensuring that files and software have not been tampered with. The EAC's VEPBCR 1.0 (April 8, 2024) references NIST SP 800-53 rev. 5 (SI-7), requiring software and data integrity verification through cryptographic mechanisms.
59. Texas Election Code §129.023(c-1) mandates hash validation for voting systems. The Secretary of State issued Advisory 2022-30 implementing this requirement. However, the Secretary has not extended this requirement to electronic pollbooks. The KnowInk PollPad does not provide hash validation of its software or data files. If hash validation had been required and implemented, the anomalous records would have been immediately detectable.
S. Harris County Precedent
60. Dr. Paquette's prior peer-reviewed study -- "Evidence of Sophisticated ID Assignment Algorithm in Harris County Voter Registration System" -documented a non-random ID assignment function embedded in the official voter registration numbering system of Harris County, Texas. The mathematical signatures are consistent across both datasets -- uniform spacing, integer constraint systems, and floating-point arithmetic fingerprints. The Bexar County dataset demonstrates what Dr. Paquette characterizes as the operational phase of the architecture first identified in Harris County.
T. Multi-State Algorithmic Pattern
61. Dr. Paquette and allied researchers have traced similar algorithmic artifacts in voter registration ID numbering systems across multiple states, including New York, Arizona, Georgia, and Texas. The Texas findings are unique because, for the first time, the algorithm was captured live in transit -- during the active voting window -- rather than inferred retrospectively from voter roll analysis.
VI. ANTICIPATORY REBUTTAL: WHY ACCIDENTAL CAUSES ARE EXCLUDED
62. Respondents may contend that the fractional State IDs are merely an "Excel formatting artifact" or a "display error" that arose during data export. This section addresses and refutes that contention in advance, because the compliance question under Title III is not whether the formatting looks different but whether the statewide list is functioning as the single, uniform, official list required by statute.
A. The Excel Artifact Theory Fails on Multiple Independent Grounds
63. Excel does not generate fractional values from whole numbers. It may truncate long integers to apparent precision limits, but truncation removes trailing digits -- it does not add five decimal places of fractional content to a ten-digit integer. The fractional components in the Bexar County data (.79993, .57755, .66981, etc.) are not truncation artifacts. They are computed values.
64. Even if some hypothetical export pipeline introduced formatting noise, no such pipeline produces: (a) a perfectly uniform inter-record gap of 22,084.82189 across 4,109 consecutive intervals; (b) a total span that divides by that gap to yield exactly 4,109.0000 with zero remainder; (c) a palindromic four-value cycle consistent with IEEE 754 floating-point arithmetic; (d) boundary anomalies at exact multiples of 735; (e) six sequential alphabetical passes through a fixed anchor list; (f) records that occupy a 90-million-unit dead zone in the statewide ID space; and (g) two distinct address fabrication strategies applied systematically across 735 groups. The simultaneous presence of all seven features excludes every accidental explanation.
B. The Compliance Question Is Not Cosmetic
65. Even if Respondents characterize the fractional IDs as a display issue, the Title III question remains: can the State produce a reliable, auditable mapping between
every poll-place check-in and the official statewide list? If the identifiers used in official check-in outputs cannot be reconciled one-to-one with valid statewide registration records, the statewide list is not functioning as the "single, uniform, official" list required by 52 U.S.C. § 21083, regardless of the cause.
66. The burden is on the State to demonstrate reconcilability, not on the Complainant to prove the mechanism. Title III requires a system that works. A system that produces thousands of non-reconcilable identifiers -- whether through malfunction or manipulation -- does not work.
C. The Glitch Cannot Solve Integer Systems
67. The endpoints of the fractional sequence -- Abel at position 1 and Braswell at position 435 -- are selected by the same mechanism: last-name alphabetical sort applied to two different positions, each determined by a different mathematical constraint. Position 1 is the beginning of the alphabet. Position 435 is the unique solution to the integer constraint system that determines the 5-copy / 6-copy boundary. A glitch does not solve two-equation integer systems. A glitch does not derive a unique stopping position from that solution. A glitch does not sort a voter list alphabetically and assign calculated minimum and maximum values to positions 1 and 435. Those are the outputs of a specification, not a malfunction.
VII. VIOLATIONS OF TITLE III OF THE HELP AMERICA VOTE ACT
Pursuant to 1 TAC §81.171(c)(3), Complainant alleges the following violations. Each identifies the responsible Respondents and the statutory basis. Where the term "anomalous records" is used, it refers to the 4,110 fractional-ID records described in the Statement of Facts. Where the term "system" is used, it refers to the KnowInk electronic pollbook system deployed in Bexar County.
VIOLATION 1: Failure to Maintain Single, Uniform, Official Statewide List
HAVA Section 52 U.S.C. § 21083(a)(1)(A) (HAVA §303(a)(1)(A))
Requirement Each State shall implement and maintain a single, uniform, official, centralized, interactive computerized statewide voter registration list that serves as the official voter registration list
Administrative Complaint -- Bernegger v. Carew et al. for the conduct of all federal elections.
Finding The KnowInk system operated with 4,110 records bearing State IDs that do not exist in the Texas TEAM database and, per Dr. Paquette, never entered the county or state voter registration lists. These records existed only within the KnowInk system, entirely bypassing the official statewide list. When a county check-in system presents, accepts, or processes records that cannot be reconciled to the official statewide list identifier, the State is no longer operating a functionally single and uniform official list at the pollingplace interface. The use of identifiers not found in the official statewide list defeats the requirement of uniformity. The replication of 735 valid voters into 4,110 derivative identifiers demonstrates structural fragmentation of the voter list.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence Forensic Report v20, Dataset Description; statewide database verification; Exhibits A, C
VIOLATION 2: Failure to Eliminate Duplicate Registrations
HAVA Section 52 U.S.C. § 21083(a)(2)(B)(iii) (HAVA §303(a)(2)(B)(iii))
Requirement The computerized list shall be maintained with provisions to ensure that duplicate names are eliminated from the official list.
Finding 735 voters were duplicated 5-6 times each, generating 4,110 derivative records. These are not clerical duplicates of whole numbers but algorithmically spaced fractional derivatives. The creation of artificial duplicate voter entries is the opposite of the statutory mandate to eliminate duplicate registrations. The KnowInk system's 15-minute anti-doublevoting synchronization either failed or was bypassed. No list maintenance procedure detected these duplicates before,
HAVA Administrative Complaint -- Bernegger v. Carew et al. during, or after voting.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence Forensic Report v20, Clone Group Structure; Exhibit C; Precinct 3075 spreadsheet
VIOLATION 3: Failure to Provide Adequate Technological Security
HAVA Section 52 U.S.C. § 21083(a)(3) (HAVA §303(a)(3))
Requirement The State or local official shall provide adequate technological security measures to prevent unauthorized access to the computerized list.
Finding 4,110 anomalous records were introduced into the KnowInk system. The operation required write access, reconnaissance of the statewide ID space, and pre-harvesting of genuine voter identities. The KnowInk system's active network connectivity during elections created the attack surface. The system's user interface conceals fractional ID components from poll workers. The anomalous records include a built-in deletion mechanism: a single database query filtering for noninteger IDs returns exactly 4,110 records for batch removal. Respondents Carew and Bexar County Elections Department failed to prevent unauthorized access. The Secretary certified the system without requiring injection-resistance testing or hash validation.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence Forensic Report v20, Logical Proof, Fraud Mechanism; v18, Palindromic Gap; Martinez Press Release; certification records
VIOLATION 4: Failure to Ensure Accurate Records
HAVA Section 52 U.S.C. § 21083(a)(4)(A)-(B) (HAVA §303(a)(4)(A)-(B))
Requirement The State shall ensure that voter registration records in the computerized list are accurate and are updated regularly, and shall include safeguards to ensure that eligible voters are not removed in error.
Finding The 4,110 fractional-ID records are demonstrably inaccurate: they reference identifiers that do not exist in any legitimate database. Their introduction replaced or displaced in the exported check-in dataset - conceivably overwroteapproximately 4,110 legitimate D-through-Z voter records from the February 18 check-in file. These real voters who checked in on February 18 have no surviving record in the production file system. The KnowInk system permitted this data overwriting without integrity checks, without alerts, and without preserving the original records.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence
Forensic Report v20, Three-File Timeline; alphabetical distribution analysis
VIOLATION 5: Failure to Verify Voter Registration Information
HAVA Section 52 U.S.C. § 21083(a)(5)(A) (HAVA §303(a)(5)(A))
Requirement The State must verify an applicant's driver's license number or the last four digits of the applicant's social security number before the registration is accepted.
Finding
The 4,110 anomalous records were never processed through identity verification. Their State IDs occupy a void in the statewide ID space where no registrations exist. They never entered the county or state voter registration lists. Yet the KnowInk system accepted and processed check-ins for these unverified records without generating any alert or requiring provisional ballot procedures.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence
Forensic Report v20, ID Space Analysis; statewide database verification
VIOLATION 6: Failure to Administer Provisional Voting Safeguards
HAVA Section 52 U.S.C. § 21082(a) (HAVA §302(a))
Requirement If an individual's name does not appear on the official voter registration list, the individual shall be permitted to cast a provisional ballot subject to later verification.
Finding The anomalous records reference identifiers not present in the official statewide list. Under HAVA §302, any check-in that cannot be verified against the statewide list must be routed through provisional ballot procedures. The KnowInk system permitted these records to generate regular (not provisional) ballot authorizations, bypassing the federally required safeguard. Additionally, legitimate D-through-Z voters whose records were overwritten may have been denied voting or improperly routed when their records could not be found.
Respondents Respondents Carew and Bexar County Elections Department
Evidence
Forensic Report v20, Three-File Timeline; ballot generation analysis
VIOLATION 7: Failure to Maintain Audit Capacity and Permanent Records
HAVA Section 52 U.S.C. § 21081(a)(2)(A)-(B) (HAVA §301(a)(2)(A)-(B))
Requirement The voting system shall produce a permanent paper record with a manual audit capacity for such system.
Finding The February 25 bulk replacement of all 14 CSV files in a 78-second automated operation destroyed the original audit trail. The chain of custody is permanently broken. No current
Respondents
Evidence
HAVA Administrative Complaint -- Bernegger v. Carew et al.
file can be certified for recount or reconciliation without independent physical canvass of every voter on the check-in record. A system in which thousands of voter check-ins can be created and then deleted without public reconciliation fails minimum audit integrity requirements. Whether the cause is malfunction or manipulation is immaterial: in either case, the system is not auditable.
Respondents Carew, Bexar County Elections Department, and Nelson
Forensic Report v20 Appendix; filesystem timestamps; Exhibit D
VIOLATION 8: Failure to Maintain Records of System Errors
HAVA Section 52 U.S.C. § 21081(b)(2)(C)-(D) (HAVA §301(b)(2)(C)-(D))
Requirement The voting system definition includes practices and documentation to maintain records of system errors and defects.
Finding No error logs, incident reports, or change-control records have been disclosed for the anomalous records or the bulk file replacement. The Secretary has not required production of KnowInk server-side logs, cloud database records, or ePlus control station data.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence
Martinez Press Release (Feb. 26); absence of responsive disclosure
VIOLATION 9: Failure to Permit Voter Verification
HAVA Section 52 U.S.C. § 21081(a)(1)(A)(i)-(ii) (HAVA §301(a)(1)(A)(i)(ii))
Requirement The voting system shall permit the voter to verify the votes selected and make changes before the ballot is cast and counted. Epoll pads/books are to follow the EAC’s Best Practices, which the State of Texas claims they adhere to.
Finding If the 4,110 anomalous records generated ballots, those ballots were cast without any legitimate voter present to verify or authorize them. No real voter authorized these ballot transactions. Each affected voter confirmed voting only once. The Sec. of State of Texas nor Bexar County election officials, their “certified” or “approved” epoll pads adhere to the Best Practices.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence Forensic Report v20, Ballot Issuance; Martinez public address (Feb. 24)
VIOLATION 10: Failure to Ensure Security Standards in Certified Systems
HAVA Section 52 U.S.C. § 21081(a)(3) (HAVA §301(a)(3))
Requirement The State shall adopt voting system standards that meet or exceed the requirements of §301(a).
Finding The Secretary certified the KnowInk system based on Pro V&V reports that did not test for injection resistance, fractional ID detection, anti-double-voting bypass, audit trail immutability, or hash validation. Pro V&V's prior certification of the subsequently-decertified ES&S ExpressPoll demonstrates a pattern of inadequate testing. The certification standards were insufficient to ensure HAVA §301(a) compliance. The Sec. of State of Texas nor Bexar County election officials, their “certified” or “approved” epoll pads, adhere to the Best Practices.
Respondents Respondent Nelson
Evidence Certification letters (Jan. 2020, Aug. 2024, Mar. 2025, Oct.
Administrative Complaint -- Bernegger v. Carew et al. 2025); ES&S decertification (Dec. 2024)
VIOLATION 11: Failure to Implement Hash Validation
HAVA Section 52 U.S.C. § 21083(a)(3) and § 21081(a)(2) (HAVA §303(a) (3), §301(a)(2))
Requirement Adequate technological security measures (§303(a)(3)) and permanent record with audit capacity (§301(a)(2)).
Finding The KnowInk PollPad does not provide hash validation of its software or data files. The EAC's VEPBCR 1.0 (referencing NIST SP 800-53 SI-7) and the EAC's own publication identify hash validation as critical. Texas Election Code §129.023(c-1) mandates hash validation for voting systems. The Secretary has not extended this requirement to electronic pollbooks. If hash validation had been required, the anomalous records would have been immediately detectable.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence EAC VEPBCR 1.0; EAC Blog (July 12, 2021); Texas SOS Advisory 2022-30; TEC §129.023(c-1)
VIOLATION 12: Use of Network-Connected Pollbooks During Federal Elections
HAVA Section 52 U.S.C. § 21083(a)(3) (HAVA §303(a)(3))
Requirement Adequate technological security measures to prevent unauthorized access.
Finding The KnowInk Poll Pad devices actively connect to communications networks during federal elections, transmitting and receiving voter data via cellular and Wi-Fi connections every 15 minutes. This real-time connectivity creates a continuous attack surface. A system that must connect to external networks to function is inherently
HAVA Administrative Complaint -- Bernegger v. Carew et al.
vulnerable to the type of data anomaly documented in this complaint. The forensic evidence demonstrates that this attack surface was in fact exploited or that it failed to prevent the introduction of anomalous data. The Sec. of State of Texas nor Bexar County election officials, their “certified” or “approved” epoll pads, adhere to the Best Practices.
Respondents Respondents Carew, Bexar County Elections Department, and Nelson
Evidence
KnowInk system architecture documentation; Forensic Report v20
VIOLATION 13: Failure to Provide Appropriate Remedy
HAVA Section 52 U.S.C. § 15512(a)(2)(F) (HAVA §402(a)(2)(F))
Requirement If the State determines there is a violation, the State shall provide the appropriate remedy.
Finding Formal complaints were filed February 26, 2026. The bulk file replacement occurred on February 25 -- during active early voting -- raising spoliation concerns. The Secretary has not acted to suspend or decertify the KnowInk system pending investigation. No explanation for the anomalous data or the bulk replacement has been provided.
Respondents Respondent Nelson (primary); all Respondents (preservation)
Evidence Martinez Press Release; bulk replacement timeline; Exhibit D
VIII. STATE PLAN COMPLIANCE FAILURE
68. Texas adopted its HAVA State Plan in January 2005. That Plan was developed under Section 254 of HAVA and constitutes the State's commitment to the federal government regarding how it would implement Title III requirements.
69. Electronic poll pads -- including the KnowInk system deployed in Bexar County -- were implemented long after the State Plan was adopted. The State Plan
contains no provision describing, authorizing, or imposing safeguards on: (a) the use of electronic pollbook systems that generate, accept, or process voter identifiers not present in the official statewide list; (b) the algorithmic generation of voter identifiers by pollbook software; (c) audit log requirements for batch file replacement operations; (d) hash validation requirements for pollbook data; or (e) network connectivity standards for pollbook devices during active voting.
70. Section 254(a)(11) of HAVA (52 U.S.C. § 21004(a)(11)) provides that the State "may not make any material change in the administration of the plan" unless the change is developed and published in the Federal Register, subject to public notice and comment, and takes effect only after the expiration of a 30-day period. The introduction of network-connected electronic pollbooks that replace the traditional paper-based check-in process constitutes a material change to the mechanism by which voter identity is verified in federal elections.
71. Upon information and belief, no amendment to the Texas State Plan has been published in the Federal Register addressing the adoption of electronic pollbooks, the certification standards applicable to such systems, or the audit and security controls required for their operation. The absence of a State Plan amendment addressing electronic pollbook controls raises compliance concerns under 52 U.S.C. § 21004 and undermines the State's commitment under Section 254 to maintain uniform and transparent election administration procedures.
IX. REQUESTED FINDINGS AND PRODUCTION
72. Complainant respectfully requests that the Secretary of State issue written findings determining whether:
(a) The 4,110 fractional State IDs existed in the live early voting environment during the February 2026 early voting period;
(b) The constant inter-record interval of 22,084.82189 was produced by software logic;
(c) The fourteen CSV files were replaced in a synchronized batch operation on February 25, 2026;
(d) The 735 integer anchor IDs were replicated algorithmically;
(e) The KnowInk system is capable of producing, accepting, or processing voter identifiers not present in the official statewide list;
(f) The Bexar County check-in records at issue are fully reconcilable to the official statewide voter registration list.
73. To make these determinations, the Secretary should require production of:
(a) The original source-system exports for the relevant election period, including the pre-replacement versions of all 14 CSV files;
(b) System audit logs reflecting ID generation activity, check-in events, and data synchronization;
(c) KnowInk server-side data, cloud database records, and ePlus control station logs;
(d) Vendor software version logs during the relevant period;
(e) The applicable data dictionary defining all identifier fields, including "State ID";
(f) Pro V&V test reports and testing protocols for all four KnowInk certification submissions;
(g) Reconciliation reports demonstrating one-to-one correspondence between check-ins and valid statewide registrations.
74. These materials are necessary not as a general investigation but to permit a specific Title III compliance finding as to list uniformity, identifier integrity, provisional-ballot safeguards, and auditability. Without such documentation, a determination of compliance under 52 U.S.C. §§ 21081-21083 cannot be meaningfully made on the record.
X. REMEDY REQUESTED
Complainant respectfully requests the following remedies pursuant to HAVA §402(a)(2)(F) and 1 TAC §81.171(f):
1. A finding of violation for each of the thirteen HAVA Title III violations identified in Section VII, against all named Respondents as specified.
2. An immediate order directing all Respondents to preserve all election data, records, logs, poll pad data, ePlus control station data, cloud database records, server-side data, synchronization logs, access logs, testing reports, certification submissions, ballot images, chain of custody documentation, text and email communications, and all other materials related to the March 2026 Republican Primary in Bexar County.
3. An order directing Respondents Carew and Bexar County Elections Department to provide a written explanation for: (a) the 4,110 fractional State ID records; and (b) the February 25 bulk file replacement.
4. An order directing a full forensic audit of all Bexar County early voting data, including analysis of KnowInk system logs from all components (Poll Pads, ePlus, cloud database), and a physical canvass of every voter on the check-in record.
5. An order directing the Secretary of State to conduct a statewide investigation into the integrity of KnowInk systems and all electronic pollbook systems, and to review all certifications under Texas Election Code §31.014 and §122.001, including the adequacy of Pro V&V's testing protocols. This investigation should specifically encompass Harris County, where Dr. Paquette's prior study documented non-random algorithmic ID assignment patterns.
6. Immediate suspension or decertification of the KnowInk PollPad system pending completion of the investigation and independent security audit.
7. A ban on the use of electronic pollbooks that maintain active network connectivity during the conduct of federal elections in the State of Texas, or, in the alternative, an order requiring air-gapped operation during voting, with data transfer occurring only through physically secured, hash-validated media.
8. An order requiring hash validation for all electronic pollbook data files -including check-in records, voter registration data, and software -- as a mandatory condition of certification under Texas Election Code §31.014, consistent with the requirements already imposed on voting systems under §129.023(c-1).
9. A halt to certification of the Bexar County early voting results pending investigation and audit.
10. Referral to the Texas Attorney General for investigation under state law, including the Texas Election Code.
11. Referral to the U.S. Department of Justice, Civil Rights Division, Voting Section, for enforcement under HAVA §401 (52 U.S.C. §15511), with notice of the multi-state algorithmic pattern documented in New York, Arizona, Georgia, and Texas.
12. If intentional conduct is established, referral for criminal investigation under HAVA §905 (52 U.S.C. §15544) and applicable state criminal statutes.
13. An order directing strengthened electronic pollbook certification standards requiring: injection resistance testing, audit trail immutability, back-end access controls, anti-double-voting enforcement, hash validation of all data files and software, and air-gapped or cryptographically secured network protocols.
14. An order requiring that all electronic pollbook systems certified for use in Texas implement database-level validation rules enforcing integer-only State Voter ID fields, rejecting any record with a fractional or non-integer ID value at the point of entry, and generating an immediate alert upon any attempt to load non-integer ID data.
15. An order requiring public disclosure of which specific requirements of the Texas Technical Testing Matrix and Texas Electronic Pollbook Functional Standards the KnowInk PollPad system does and does not currently meet.
16. A determination regarding the State Plan's compliance with 52 U.S.C. § 21004(a)(11) as it relates to the adoption of electronic pollbook systems without published State Plan amendment.
17. Any other appropriate remedy to protect the integrity of the March 3, 2026 Primary Election and future elections.
Complainant acknowledges that, under 1 TAC §81.171(f), the remedy may not include monetary damages, costs, attorney fees, election invalidation, or ballot validity determinations.
XI. LIST OF SUPPORTING EXHIBITS
The following exhibits are attached hereto and incorporated by reference:
Exhibit A: Constant Interval Demonstration Table -- 14 consecutive fractional State IDs showing identical inter-record gap of 22,084.82189
Exhibit B: Four-Value Gap Cycle Table -- Complete distribution of gap values at 10-digit precision with IEEE 754 analysis
Exhibit C: Anchor Replication Table -- Representative sample of 735 anchor voters with clone counts and address patterns
Exhibit D: Bulk File Replacement Timeline -- All 14 CSV files with creation and modification timestamps
Exhibit E: Spreadsheet screenshot and data: check-in file showing integer IDs and fractional clones, including illustrative examples
Exhibit F: Bexar County Forensic Analysis, Version 20, by Dr. Andrew Paquette (February 27, 2026)
Exhibit G: Bexar County Forensic Analysis, Version 18, by Dr. Andrew Paquette
Exhibit H: Martinez Campaign Press Release (February 26, 2026)
Exhibit I: Weston Martinez Public Address (February 24, 2026, Alamo Plaza)
Exhibit J: Original early voting check-in file (TOTAL REPUBLICAN VOTER CHECK IN AS OF 02 18 2026.xlsx)
Exhibit K: Replacement files from February 25, 2026 with filesystem timestamp documentation
Exhibit L: Dr. Paquette audio presentation and transcript, "The Weston (Alamo) Algorithm"
Exhibit M: Texas Secretary of State KnowInk certification letters: v2.4.9, v3.6, v3.6.3, v4.1.3
Exhibit N: Texas SOS Electronic Pollbook Certification page and functional standards
Exhibit O: Texas SOS ES&S ExpressPoll decertification letter (December 2024)
Exhibit P: EAC Voluntary Electronic Poll Book Certification Requirements (VEPBCR 1.0), April 8, 2024
Exhibit Q: EAC publication: "What is Hash Validation and Why Should Election Officials Care?" (July 12, 2021)
Exhibit R: Texas SOS Advisory 2022-30: Hash Validation Procedures
Exhibit S: Bexar County Commissioners Court records (May 13, 2025) re: KnowInk procurement
Exhibit T: KnowInk system architecture and marketing materials
Exhibit U: Dr. Andrew Paquette, "Evidence of Sophisticated ID Assignment Algorithm in Harris County Voter Registration System"
Exhibit V: Multi-state algorithmic voter ID pattern analyses (New York, Arizona, Georgia, Texas)
Exhibit W: Texas HAVA State Plan, January 2005
Exhibit X: Help America Vote Act of 2002, P.L. 107-252 (relevant sections)
XII. VERIFICATION, OATH, AND SIGNATURE
Pursuant to Section 402(a)(2)(C) of HAVA and 1 TAC §81.171(c)(1), this complaint is submitted in writing, signed, and notarized by the Complainant, who swears under oath that the facts alleged herein are true and correct to the best of Complainant's knowledge and belief.
Peter Bernegger, Complainant
Date: _________________________
Full Name: Peter Bernegger
Telephone:
Mailing Address:
Before me, the undersigned notary public, on this _____ day of March, 2026, personally appeared Peter Bernegger, known to me (or proved to me on the basis of satisfactory evidence) to be the person whose name is subscribed to the within instrument and acknowledged to me that he executed the same in his authorized capacity, and that by his signature on the instrument the person, or the entity upon behalf of which the person acted, executed the instrument.
SUBSCRIBED AND SWORN TO before me this _____ day of _____________, 2026.
Notary Public, State of Wisconsin
My commission expires: ___________________
[NOTARY SEAL]
EXHIBIT A
Constant Interval Demonstration Table (14 Consecutive Fractional State IDs with Calculated Differences)
The following table lists 14 consecutive records from the 4,110 fractional-ID dataset, sorted in ascending order by State ID. The "Difference" column shows the gap between each record and the next. Every difference is identical: 22,084.82189. This uniformity across 4,109 consecutive intervals, producing a total span that divides to yield exactly 4,109.0000 with zero remainder, is the mathematical signature of deliberate algorithmic construction.
EXHIBIT B
Four-Value Gap Cycle Analysis
(IEEE 754 Floating-Point Arithmetic Fingerprint)
At full 10-digit decimal precision, the 4,109 inter-record gaps resolve into exactly four distinct values. These four values form two conjugate micro-pairs consistent with IEEE 754 double-precision floating-point arithmetic, in which the same division operation executed in forward versus reverse order produces rounding residuals at the 10th decimal place. The values cycle in palindromic seven-element blocks (A B A B C D C / C D C B A B A), with boundary anomalies at exact multiples of 735 -- independently confirming the anchor group size.
Gap Value (10-digit precision)
The palindromic structure cannot arise from manual data entry, database migration, display formatting, software misconfiguration, or random corruption. It is the deterministic output of compiled software executing a specific floating-point loop in alternating directional passes over a fixed-size record set.
EXHIBIT C
Anchor Replication Table
(Representative Sample of 735 Anchor Voters and Clone Patterns)
Each of the 4,110 fractional records corresponds to one of 735 genuine registered voters. The table below shows a representative sample. Of the 735 groups, 300 received 5 clones and 435 received 6 clones (the unique integer solution yielding 4,110 total). Address patterns divide into two categories: Sequential +1 (67% of groups) where house numbers increment by one per clone, and Identical (33%) where all clones share the anchor's address. Only 2 of 735 groups deviate from these two patterns.
Anchor Voter (Name)
Integer State ID # Clones Address Pattern
Robert Earl Abel 1068101234 6
Maria L. Acosta 1071255678 5
James R. Adams 1073389012 6
Linda K. Allen 1075523456 5
David M. Alvarez 1077657890 6
Patricia A. Anderson 1079791234 5
Michael J. Bailey 1081925678 6
Sarah E. Baker 1084059012 5
Thomas W. Barnes 1086193456 6
Sequential +1 (101, 102, 103...)
Sequential +1 (205, 206, 207...)
Identical (same address)
Sequential +1 (412, 413, 414...)
Sequential +1 (1501, 1502, 1503...)
Identical (same address)
Sequential +1 (800, 801, 802...)
Sequential +1 (2100, 2101, 2102...)
Identical (same address)
Jennifer L. Bell 1088327890 5 Sequential +1 (330, 331, 332...)
Note: Names shown are illustrative of the pattern structure. 99.5% of anchor surnames begin with A, B, or C -- a statistical impossibility for organic voting data, where A/B/C surnames constitute approximately 18.7% of the general population. The anchor list terminates at "Chicas" -- the exact alphabetical cutoff where the counter reached 735.
EXHIBIT D
Bulk File Replacement Timeline (February 25, 2026 -- Synchronized 78-Second Operation)
On February 25, 2026, all 14 CSV early voting files were simultaneously replaced. The creation and modification timestamps below demonstrate that all 14 files were written in a single automated batch operation lasting 78 seconds. Post-replacement, every file contains zero fractional State IDs. The original data containing the anomalous records survives only because it was captured on February 19, 2026 and distributed to precinct chairs before the replacement occurred.
EV_Checkins_REP_02172026.csv 08:46:26 08:47:44 0
EV_Checkins_REP_02182026.csv 08:46:26 08:47:44 0
EV_Checkins_REP_02192026.csv 08:46:26 08:47:44 0
EV_Checkins_REP_02202026.csv 08:46:26 08:47:44 0
EV_Checkins_REP_02212026.csv 08:46:26 08:47:44 0
EV_Checkins_REP_02222026.csv 08:46:26 08:47:44 0
EV_Checkins_REP_02232026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02172026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02182026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02192026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02202026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02212026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02222026.csv 08:46:26 08:47:44 0
EV_Checkins_DEM_02232026.csv 08:46:26 08:47:44 0
All timestamps are on February 25, 2026. Creation time for all 14 files: 08:46:26. Modification time for all 14 files: 08:47:44. Total elapsed time: 78 seconds. This synchronization across 14 files (7 Republican, 7 Democratic, spanning 7 dates each) is consistent with an automated batch operation, not manual file-by-file editing.
MATHEMATICAL APPENDIX
Why Constant-Interval Sequences Cannot Arise from Random Corruption
This appendix explains, for the non-technical reader, why the constant-interval structure in the Bexar County data excludes all accidental explanations.
1. The Nature of Random Corruption
When electronic systems experience data corruption -- from transmission errors, storage degradation, software bugs, or formatting artifacts -- the resulting errors are either random (affecting different values in different ways) or systematic in a simple, uniform manner (such as truncating all numbers to a fixed length). Random corruption does not produce structured mathematical relationships between corrupted values. It produces noise.
2. The Constant-Interval Test
Consider 4,110 numbers. If they are randomly generated or randomly corrupted, the gaps between consecutive sorted values will vary. The probability that all 4,109 gaps will be identical to five or more decimal places is, for practical purposes, zero. It is equivalent to rolling a ten-sided die 4,109 times and getting the same five-digit number every time.
3. The Perfect Division Test
Even more powerfully: if we take the total span of the 4,110 numbers (the largest minus the smallest) and divide by the constant gap, we should get exactly 4,109 -because 4,110 numbers produce 4,109 intervals. The Bexar County data yields exactly 4,109.0000 with zero remainder. This is not a statistical argument. It is arithmetic. The only way to produce this result is to construct the sequence deliberately: choose a starting point, choose a gap, and generate each successive value by adding the gap to the previous value.
4. The Compounding Impossibility
The constant-interval structure is only one of at least seven independent features that must all be present simultaneously. Each feature individually excludes
accidental causation. Together, they constitute an interlocking algebraic system in which every number is the output of a prior calculation. No class of system failure produces interlocking algebraic systems. Software specifications produce interlocking algebraic systems.