Safeguard American Voter Eligibility Act (SAVE America Act)
H.R.7296 / S.1383 — 119th Congress
Bill: H.R.7296 / S.1383 — 119th Congress
Track: Legislative
Status: House passed February 11, 2026 (218–213). Pending Senate vote. Requires 60 votes under current filibuster rules.
Lead sponsors: Sen. Mike Lee (R-UT), Sen. Lindsey Graham (R-SC)
Amends: National Voter Registration Act of 1993 (NVRA), 52 U.S.C. §20501 et seq.; Help America Vote Act (HAVA), 52 U.S.C. §21083
Plain-language summary
The SAVE America Act amends the National Voter Registration Act to require anyone registering to vote in federal elections to provide documentary proof of U.S. citizenship — in person, at a government office — at the time of registration or any subsequent registration update. Acceptable documents include a U.S. passport, birth certificate, certificate of citizenship, or a REAL ID-compliant ID that specifically indicates citizenship status. REAL ID is listed as qualifying, but in practice most states’ REAL IDs do not indicate citizenship and can be issued to legally residing noncitizens, limiting its practical utility under the statute.
The bill also requires a government-issued photo ID indicating citizenship to cast a ballot in federal elections, or an ID paired with supplemental documentation establishing citizenship. Mail-in voters must submit a copy of qualifying photo ID both with their absentee request and with their returned ballot. States are required to use federal databases including DHS’s SAVE program to identify and remove noncitizen registrants, and must remove voters from active rolls at any time upon receipt of “documentation or verified information” of noncitizenship — a standard the bill does not define. Criminal penalties and private right of action attach to election officials who register applicants without verified documentation. The bill takes effect immediately upon signing, without a defined transition or phased implementation period.
Verdict: This bill addresses a legitimate structural interest — ensuring only citizens are registered to vote in federal elections — through a mechanism whose predicted harm to eligible citizens is measurably larger than the problem it solves, and whose implementation design is so technically defective that it would produce inconsistent, ungovernable administration regardless of intent.
Problem scale vs. solution scale
Approximately 21 million eligible citizens lack readily available citizenship documents, according to the Brennan Center for Justice. USCIS voter verification data shows that only 0.04% of verification cases are returned as noncitizens — and some of those are false positives. The Heritage Foundation’s database of confirmed cases, drawn from across the political spectrum, shows fewer than 100 confirmed noncitizen ballot castings nationwide between 2000 and 2025.
The Kansas precedent is the closest analog to what this bill would produce at scale. When Kansas implemented a state-level documentary proof requirement, noncitizen registration was approximately 0.002% of registered voters. The requirement blocked roughly 31,000 eligible citizens — about 12% of all applicants — from registering. That law was later struck down for violating the NVRA. Utah’s 2025–2026 review of over 2 million registered voters found one confirmed noncitizen registration and zero instances of noncitizen voting.
A Bipartisan Policy Center analysis found that approximately 12% of registered voters do not have ready access to the documents the bill requires.
Legal impact assessment
This section is a vulnerability audit, not a litigation forecast. The question is where the structural weaknesses are, not whether courts will act on them. Citations verified through LegesGPT analysis of the bill text.
Federalism and the Elections Clause — Contested
Congress asserts authority under Article I, §4 (the Elections Clause) to require documentary proof of citizenship as a condition of registering to vote in federal elections. The structural vulnerability is whether that requirement is a permissible “manner” regulation or an impermissible federal setting of voter qualifications — which the Constitution ties primarily to state law under Article I, §2, cl. 1 and the Tenth Amendment.
The controlling precedent cuts in both directions. Arizona v. Inter Tribal Council of Arizona, 570 U.S. 1 (2013), held that the NVRA’s requirements preempted Arizona’s state-level documentary proof requirement, affirming Congress’s broad Elections Clause power over federal election registration mechanics. The bill uses that same power to impose documentary proof nationwide. But challengers can argue that Inter Tribal‘s logic was about preventing extra state hurdles to a uniform federal system — not about Congress using the Elections Clause to create de facto qualification gates. United States Term Limits, Inc. v. Thornton, 514 U.S. 779 (1995), supports the view that “manner” regulation is procedural, not substantive — a distinction challengers would press here. Oregon v. Mitchell, 400 U.S. 112 (1970), shows the Court has accepted some federal control over voting-related rules for federal elections, but the doctrine remains unsettled.
Standing is clearest for individual voters denied registration and for voter-registration organizations that can demonstrate diversion of resources — a well-established theory in election litigation. State standing via anti-commandeering is weaker here because the Elections Clause is an express delegation that can displace Tenth Amendment objections when Congress acts within it.
Fundamental right to vote — Anderson-Burdick analysis — High
The controlling framework is the Anderson-Burdick sliding scale: severe burdens on the right to vote require narrowly tailored, compelling interests; reasonable and nondiscriminatory restrictions receive more deference. Anderson v. Celebrezze, 460 U.S. 780 (1983); Burdick v. Takushi, 504 U.S. 428 (1992); Crawford v. Marion County Election Bd., 553 U.S. 181 (2008).
The bill stacks multiple burdens simultaneously. Mail-form registrants must present documentary proof in person at an election office or polling place (Sec. 2(d), new NVRA §6(e)(1)(A)–(B)) — not merely attach a copy, but make a physical trip. The blanket proof requirement applies under any method of registration (Sec. 2(b), new NVRA §4(b); Sec. 2(f), new NVRA §8(j)(1)). The photo ID requirement demands an ID that affirmatively indicates U.S. citizenship on its face (Sec. 3(c)(3)(B)); voters without such an ID must present supplemental documentation (Sec. 3(c)(4)(A)). Absentee voters must include ID copies with both their ballot request and their returned ballot (Sec. 3(b)(2)(A)–(B)). All of this takes effect immediately upon signing (Sec. 2(q); Sec. 3(e)).
Under Crawford, courts are attentive to the evidentiary record: where demonstrated fraud is rare and alternatives exist, overbroad burdens are more vulnerable. The bill’s scope — applying uniformly to all applicants including long-registered citizens making routine updates — and its lack of any implementation runway are the strongest factors in the Anderson-Burdick analysis.
Voting Rights Act Section 2 — Contested
Section 2 of the VRA, 52 U.S.C. §10301, prohibits any voting practice that “results” in denial or abridgement of the right to vote on account of race or color, assessed under the totality of circumstances. No discriminatory intent is required. The controlling modern framework is Brnovich v. Democratic National Committee, 594 U.S. 647 (2021), which articulated guideposts for Section 2 vote-denial claims: the size of the burden, the degree of racial disparity, the opportunities provided by the state’s overall voting system, the strength of state interests, and the degree of departure from 1982 practice.
The bill’s most structurally vulnerable provisions on disparate impact are the documentary proof requirement (Sec. 2(a)–(b); Sec. 2(f), new NVRA §8(j)(1)) and the citizenship-indicator ID requirement (Sec. 3(c)(3)(B)). Passport possession rates, access to certified birth certificates, and ability to obtain replacement vital records all correlate with income, age, and race. The in-person presentation requirement (Sec. 2(d), new NVRA §6(e)(1)(A)) adds transportation and time-off-work burdens that similarly correlate with socioeconomic status. Approximately 84% of women who marry change their surname, meaning roughly 69 million American women have a birth certificate that does not match their current legal name; the bill contains no provision for marriage certificates or name-change documentation, and the “other evidence” alternative process (Sec. 2(f), new NVRA §8(j)(2)(A)) leaves standards undefined. The Section 2 vulnerability is real but record-dependent under Brnovich.
Due process — voter roll removal without notice — High
The bill requires states to remove voters from active rolls “at any time” upon receipt of documentation or verified information that a registrant is not a U.S. citizen (Sec. 2(f), new NVRA §8(k)). The bill does not require pre-removal notice to the affected voter, does not define what constitutes “verified information,” and does not specify a cure timeline or appeals process.
The constitutional framework is Mathews v. Eldridge, 424 U.S. 319 (1976), which weighs the private interest at stake, the risk of erroneous deprivation, and the government’s interest. Voting is a constitutionally fundamental interest — Harper v. Virginia Board of Elections, 383 U.S. 663 (1966); Reynolds v. Sims, 377 U.S. 533 (1964) — which strengthens the private-interest factor significantly. The risk of erroneous deprivation is documented: some state audits found that up to 25% of USCIS-flagged voters had already provided citizenship documentation. Removal without notice before those errors can be caught raises due process concerns under Mullane v. Central Hanover Bank & Trust Co., 339 U.S. 306 (1950).
There is also a direct internal contradiction in federal statute. The NVRA establishes list-maintenance procedural protections including a 90-day quiet period that prohibits systematic removals close to federal elections (52 U.S.C. §20507(c)(2)(A)). The bill’s “at any time” removal language in new §8(k) directly contradicts that existing protection — it does not amend or repeal the quiet period, it simply overrides it without acknowledgment. The bill does not resolve which provision governs, meaning courts would face an irreconcilable conflict within the amended NVRA itself. This is not merely a litigation vulnerability; it is a defect in the bill’s internal consistency.
Mail voting — Constitutionally contested, operationally acute
See structural analysis below. The legal vulnerability here is compounded by timing: ballots in the 2026 midterm cycle have already been requested and in some cases mailed before any possible signing date. Retroactive application of new ID submission requirements to ballots already in transit raises both due process and equal protection concerns, and the bill contains no provision for this transition scenario.
Structural analysis
The structural question is not whether the stated goal is legitimate — it is — but whether the design achieves that goal without introducing worse problems.
This bill does not create a purely objective verification system. It creates a hybrid: a strict documentary proof requirement at the front end, coupled with a subjective discretionary fallback pathway administered by local officials, under asymmetric legal liability that penalizes approvals but not rejections. That combination does not produce rigorous, uniform verification. It produces systematic rejection bias — rational actors operating under criminal exposure will default to denying borderline applications regardless of the merits, because the cost of a false approval is personal and the cost of a false rejection is zero. The failure mode is mechanical, not political, and it operates the same way regardless of who administers the system.
Power concentration — Triggered
The bill creates a national voter database under federal executive branch control that did not previously exist. It requires all states to submit voter rolls to DHS, grants federal agencies broad discretion over verification (Sec. 2(f), new NVRA §8(j)(3)–(5)), and contains no restrictions on what DHS may do with the aggregated data, no independent auditing requirement, and no sunset or review mechanism. DHS is an executive agency whose leadership serves at the pleasure of the current president. States that have resisted prior administration demands for voter data would be compelled to comply under threat of federal enforcement. The concentration of this data in a single executive-controlled system — with no structural safeguard against partisan use — is a governance architecture problem independent of who operates it.
Architectural coupling — Triggered
Distinct from the power concentration problem is a systems design problem: the bill directly wires the voter registration system to the immigration enforcement system. It mandates that DHS investigate and potentially initiate deportation proceedings against any alien found to be unlawfully registered (Sec. 2(f), new NVRA §8(j)(5)(D)). This introduces enforcement consequences into what has historically been an administrative eligibility system — a design-level coupling that did not previously exist.
The consequences of that coupling do not depend on bad intent. Federal verification databases produce false positives at documented scale — in some state audits, up to 25% of flagged voters had already established citizenship. Under this bill’s architecture, an eligible citizen incorrectly flagged by a database error faces not just removal from voter rolls but automatic referral for immigration investigation. The system, running as designed, produces that outcome.
The bill also couples voter registration eligibility to the accuracy, availability, and interoperability of multiple federal databases — SAVE, SSA, state DMV systems — that were not designed for election administration and carry no reliability guarantees for this use case. A 24-hour response SLA baked into statute (Sec. 2(f)(5)(A)) provides no fallback when those systems fail. The eligibility determination for millions of voters is thus dependent on the real-time performance of external infrastructure the bill cannot control.
A firewall — a statutory requirement that verification findings not flow to enforcement systems without independent review — would address the coupling flaw without changing the citizenship verification goal.
Perverse incentives — Triggered
Election officials face both criminal penalties (Sec. 2(j)) and private right of action (Sec. 2(i)) for registering applicants without documentary proof. This creates a strong institutional incentive to over-reject. The downside of false negatives (eligible citizens not registered) carries no penalty; the downside of false positives (unverified applicant registered) carries criminal exposure and litigation risk. This incentive structure is not an implementation risk — it is built into the bill’s design.
Accountability gap — Triggered
The removal trigger in new §8(k) uses the phrase “verified information” without defining it. A bare database flag qualifies? A letter? A match threshold? The entire removal process — affecting potentially millions of voters — runs through an undefined standard. When eligible voters are removed on the basis of erroneous matches, there is no designated responsible party, no specified correction timeline, and no mandatory notification requirement.
Vague enforcement — Triggered
The alternative process for applicants without documentary proof (Sec. 2(f), new NVRA §8(j)(2)(A)) delegates standards simultaneously to EAC guidance and state implementation, with no minimum requirements specified in statute. The bill mandates that the process exist without defining what it must contain. This produces a patchwork of 50 different processes — maximum variation at the point where uniform protection is most needed.
This creates an unstable design triangle: a strict documentary proof requirement at the front end, a subjective discretionary fallback pathway in the middle, and criminal penalties on the official making the call at the back end. The official must decide whether an applicant’s “other evidence” sufficiently establishes citizenship — a judgment call with no defined standard — and then sign an affidavit swearing to the determination and explaining the basis for it (Sec. 2(f), new NVRA §8(j)(2)(A)(ii)–(iii)). That affidavit requirement is significant: it converts every discretionary approval into a documented, personally attributable, legally auditable decision. Every exception an official makes is on the record and tied to their name. The predictable result is systematic over-rejection of the fallback pathway regardless of how generous the EAC guidance is, because no guidance can eliminate the traceable personal liability attached to every approval.
Mail voting provisions — Triggered
Mail-in voters must submit a copy of qualifying photo ID both with their absentee request and with their returned ballot (Sec. 3(b)(2)(A)–(B)), and the ID must affirmatively indicate U.S. citizenship on its face (Sec. 3(c)(3)(B)). No current standard driver’s license indicates citizenship. The bill defines no process for what constitutes an acceptable copy, how mismatches are communicated, or what cure opportunity exists. Over 7 million Americans registered by mail in 2022; 42 states use online registration the bill would effectively eliminate. The immediate effective date means ballots already in transit are governed by rules that did not exist when they were requested.
Preemption of existing oversight — Triggered
States using back-end citizenship verification through USCIS SAVE, SSA, and DMV cross-referencing — without front-end document burdens on voters — are required to adopt the front-end documentary model. The bill simultaneously removes the NVRA’s existing procedural guardrails around list maintenance and bypasses the Paperwork Reduction Act for voter registration materials (Sec. 2(m)), eliminating a standard federal oversight mechanism for information collection burden.
Notably, the bill includes a special rule for states that do not require voter registration at all (Sec. 2(k)): those states may comply by establishing a back-end system for confirming citizenship prior to voting. This provision acknowledges that back-end verification is a functionally valid alternative — but makes it available only to the small number of states that don’t use registration, while requiring all other states to implement the front-end documentary model the bill’s own architecture implicitly concedes is not the only workable approach.
No sunset or review mechanism — Triggered
The bill contains no provision for evaluating whether it achieves its stated purpose, no trigger for reviewing disenfranchisement rates, and no expiration date on federal data aggregation. Problems that emerge in implementation have no legislated correction path.
Immediate effective date — Triggered
The bill takes effect upon signing (Sec. 2(q); Sec. 3(e)). The 2026 midterms are active with early voting already underway in multiple states. Immediate enactment creates a class of voters whose registration or ballot was valid when submitted and is invalidated by rules signed into law afterward, with no transition mechanism.
Abstraction layer analysis
This section applies a systems engineering lens to the bill’s design. Well-architected legislation defines policy goals at the statutory layer and delegates implementation specifics to rulemaking and administration — layers that can be updated without returning to Congress. This bill collapses those layers in multiple places, hardcoding implementation details that will become brittle, leaving critical interfaces undefined, and creating structural inconsistencies that cannot be corrected without new legislation.
Hardcoded document specification at the wrong layer (Sec. 2(a)(5)(A)(i)–(vii))
The bill specifies seven technical requirements for what makes a birth certificate acceptable — including that it must list “the full names of one or both parents,” carry a specific seal, and show the date it was filed with the vital records office. This level of technical specification belongs in EAC rulemaking, not statute. Amended birth certificates, tribal records that predate standardization, records from territories, and certificates issued before vital records modernization may not meet these exact requirements. When document standards evolve, correcting a statute requires Congress. Correcting a regulation requires an agency.
Technology assumption hardcoded into law (Sec. 3(b)(1)(A))
In-person voters must present a “tangible (not digital) document.” Mobile driver’s licenses (mDLs) are being standardized under ISO 18013-5 and are being adopted by states now. This bakes a 2026 technology assumption into statute. Within years this provision will directly conflict with document infrastructure the federal government is simultaneously encouraging. This is the legislative equivalent of hardcoding a file format into an API.
A past date that creates a permanent, uncorrectable two-tier system (Sec. 3(c)(4)(B)(ii)(I))
States that have submitted voter rolls to SAVE quarterly “since June 1, 2025” qualify for an exemption from the citizenship-indicator ID requirement for existing registrants. That date is already in the past. States that were not complying before this bill was written can never retroactively qualify, regardless of what they do going forward. Which rules apply to a voter is permanently determined by past state compliance, with no correction mechanism. The correct implementation is a rolling compliance window, not a fixed historical anchor.
Operational service-level agreement baked into statute (Sec. 2(f)(5)(A))
Federal agencies must respond to state verification requests within 24 hours. That is a service-level agreement — the kind of operational commitment that belongs in an interagency agreement, not a statute. There is no exception for system downtime, no fallback process, no escalation path, and no defined consequence for noncompliance. When systems fail, the statute is silent.
Ten-day implementation guidance window (Sec. 2(l))
The EAC must issue guidance to all 50 states within 10 days of enactment. For a bill that changes the mechanics of voter registration in every state simultaneously, a 10-day window guarantees the guidance will be incomplete. The provision conflates issuing guidance with issuing adequate guidance. Incomplete guidance, issued under a statutory deadline, then becomes the basis for state implementations that will vary widely and cannot easily be corrected.
Undefined interface at the critical removal trigger (Sec. 2(f), new §8(k))
The voter roll removal trigger — “documentation or verified information that a registrant is not a United States citizen” — leaves “verified information” entirely undefined. This is the most consequential undefined interface in the bill. Does a bare database flag qualify? A letter? A match above some confidence threshold? The entire due process vulnerability identified in the legal section flows from this undefined interface. In systems terms: this is a function call with no defined contract for what constitutes valid input.
Liability placed at the point of maximum interface ambiguity (Sec. 2(i)–(j))
Criminal penalties and private right of action are imposed on election officials who register applicants without documentary proof. The alternative process for applicants who cannot provide documents (Sec. 2(f)(2)(A)) is simultaneously undefined — delegated to EAC guidance and state discretion with no minimum standards specified in statute. Maximum liability is thus placed at the point where the interface between policy and implementation is most incomplete. The standard an official must meet to avoid criminal exposure depends on guidance that does not yet exist and will vary by state.
Recommendations
The stated goal — ensuring only U.S. citizens are registered to vote in federal elections — is a legitimate structural interest. The following alternatives address the same goal with substantially reduced collateral harm, better accountability architecture, and sound implementation design.
Back-end verification instead of front-end documentation burden. Back-end verification — where election officials cross-reference existing government databases (DMV, SSA, USCIS SAVE program) without requiring voters to produce physical documents — achieves the same citizenship-verification goal while affecting only applicants who cannot be confirmed through existing records. Several states are already using this model. The Kansas and Arizona front-end experience demonstrates that documentary requirements block eligible citizens at rates far exceeding the noncitizen registration rate they target.
Provisional voting and cure period for flagged voters. Any voter flagged by database verification — whether during registration or voter roll maintenance — should have the right to cast a provisional ballot and a defined cure period to produce documentation or attest citizenship. This preserves the verification function while closing the accountability gap created by the bill’s “at any time” removal with no notice requirement, and directly addresses the due process exposure identified in the legal section.
Data use restrictions and independent oversight for any federal voter roll aggregation. If federal voter roll sharing proceeds in any form, functional design requires explicit statutory restrictions on permissible uses of the data — including a firewall between voter verification and immigration enforcement — independent oversight with public reporting, defined error-correction procedures, and data security and retention standards.
Standardized mail voting ID cure process. If mail voter ID requirements are retained, the bill must define what constitutes an acceptable copy of qualifying ID, how mismatches are communicated to voters, what cure window is available, and who bears administrative responsibility. The current text imposes the requirement without any of the administrative architecture needed to implement it without mass ballot rejection.
Delegate document specifications to rulemaking, not statute. The bill’s seven-part birth certificate specification and the “tangible not digital” ID requirement should be replaced with delegation to EAC rulemaking — with required minimum standards defined in statute and a mandatory update cycle. This keeps the policy goal in law while allowing the implementation details to be maintained without returning to Congress.
Replace the June 1, 2025 compliance anchor with a rolling window. The SAVE exemption should be based on a defined compliance period prior to each election, not a fixed historical date. This allows states to qualify going forward and eliminates the permanent two-tier system the current text creates.
Phased implementation keyed to election cycles. No material change to voter registration requirements or ballot casting rules should take effect mid-election cycle. A minimum 18-month implementation window, taking effect after the 2026 midterms, would allow states to build compliant systems, train election officials, and develop adequate guidance — and allow the EAC more than 10 days to produce it.
Church Bells — The Statecraft Blueprint Legislative track brief
Sources: H.R.7296 bill text (119th Congress); Bipartisan Policy Center; Brennan Center for Justice; Heritage Foundation; Center for American Progress; Campaign Legal Center; ACLU; Issue One; The 19th; CNBC; The Conversation; Legal Talk Network.
Legal citations verified via LegesGPT.
This brief is a structural analysis, not a litigation forecast or legal advice.
Version history
v1.0 — 2026-03-19 — Initial brief, House-passed text (218–213). Senate vote pending.
v1.1 — 2026-03-19 — Legal section updated with verified citations and bill section references from LegesGPT analysis.
v1.2 — 2026-03-19 — Abstraction layer analysis added from direct bill text review.
v1.3 — 2026-03-20 — Multi-AI review incorporated: overstatements tightened, affidavit contradiction elevated, immigration enforcement coupling strengthened, NVRA internal contradiction clarified, states-without-registration edge case added.
v1.4 — 2026-03-20 — Power concentration and architectural coupling split into separate flags; coupling argument strengthened and given independent treatment.
v1.5 — 2026-03-20 — Core contradiction paragraph added naming systematic rejection bias as the central failure mode; affidavit traceable liability made explicit; architectural coupling strengthened with federal database dependency critique.
v1.6 — 2026-03-20 — Removed erroneous "Department of War" abstraction layer flag; reference is current, not stale (EO 14347, September 2025).