March 6, 2026
Maryland State Board of Elections
P.O. Box 6486
Annapolis, MD 21401
Via email
Comments on Proposed Changes to COMAR 33.08.05 Post Election Verification and Audit
Dear Members of the State Board of Elections:
On behalf of Verified Voting, I submit these comments on the proposed changes to COMAR 33.08.05 Post Election Verification and Audit, including new Regulation .09 related to the Risk-Limiting Audit. Verified Voting’s mission is to strengthen democracy for all voters by promoting the responsible use of technology in elections. Since its founding by computer scientists in 2004, Verified Voting has promoted voter-verified paper ballots and routine, rigorous post-election audits—especially risk-limiting audits—to check the accuracy of computerized voting systems.
These regulations represent a vital step in Maryland’s statewide implementation of risk-limiting audits (RLAs), first planned for after the June 2026 primary elections. An RLA offers important advantages over Maryland’s previous fixed-percentage manual audit. An RLA takes place prior to the certification of the election, and the audit provides solid evidence for the initial election outcome when it is correct. Timely sharing of RLA results can help counter attacks on the trustworthiness of the vote counts. Because RLAs provide an additional opportunity to detect and correct wrong outcomes, they also underscore election officials’ commitment to accuracy.
Overall, the draft regulations provide Maryland with flexibility as it begins to implement RLAs, as well as a path to building out its RLA program as election officials gain experience and familiarity with this new process. For instance, while the draft regulations only require one statewide contest to be audited, selected at random, they allow the State Board of Elections to select an additional number of contests for audit. Randomly selecting one or more contests is desirable because it ensures that every statewide contest has a chance of being audited. Now or in the future, the State Board may wish to consider additionally requiring that the top-of-ticket contest always be audited in elections after a certain date, and/or authorizing the State Board to select an additional contest for audit based on certain criteria or circumstances.
We also broadly agree with the choice to make batch comparison RLAs the default method for RLAs at this stage, while allowing the use of other RLA methods as needed. Other states with predominately in-person voting have also opted for batch comparison RLAs of statewide contests, including Georgia, Pennsylvania, and Rhode Island. Although batch comparison audits require examining more ballots than other methods, at this time they are substantially easier for Maryland to implement and to explain. In the future, after Maryland replaces its current voting system, it may be feasible to implement ballot-level comparison audits. When feasible, ballot-level comparison audits usually require examining many fewer ballots than batch audits, facilitating risk-limiting audits or spot checks of more contests.
We offer the following specific suggestions for changes, or at least awareness, in 2026:
- 1) We are concerned that the regulations do not explicitly enforce the basic statistical requirements of an RLA. An RLA must be able to expand until it finds sufficiently strong evidence that a reported outcome is correct (or, if necessary, the outcome is corrected). Normally, this evidential requirement can be met in a single “round” of auditing, as the draft regulation implicitly expects. Nevertheless, a valid RLA cannot simply assume that it will succeed in one round. For a statewide RLA, the State Administrator or another central authority must be charged with confirming that the RLA can conclude because the evidence satisfies the risk limit, or with requiring additional auditing or other actions if necessary. (In an RLA of a local contest under .09(C)(3), this responsibility presumably would fall to the local board.) We suggest adding a new section .09(G) (perhaps called “Completion of Risk-Limiting Audit,” parallel to (E)) to give the central authority the requisite responsibility and authority.
- 2) Relatedly, without wanting to overcomplicate the required RLA results reporting under .08(C)(1)(a), we suggest minimally adding language akin to this: “(iii) detailed audit results that demonstrate that the risk limit was satisfied.”
- 3) We can imagine scenarios in which local boards commence and complete their RLAs (i.e., their portions of the statewide RLA) on different schedules, and (as mentioned above) in which one or more local boards might be required to perform additional auditing. We suggest in .09(E), authorizing the State Administrator to hold additional public meetings if necessary in the course of the statewide RLA, for the same purposes: selecting random seeds and determining (additional) ballots to be audited. In conjunction with the proposed .09(G), we believe this would give the State Administrator sufficient authority to navigate unexpected complications in the audit.
- 4) Since the regulation identifies batch comparison RLAs as the default audit method, we suggest adding a definition for “batch tallies,” which are required for batch comparison RLAs in addition to a ballot manifest: “Batch tallies means the vote counts, reported by batch, for each candidate or choice in the audited contest(s).” Language in .09(A)(3)(a) and .09(D)(2)(a) could be revised to require local boards to prepare batch tallies in addition to a ballot manifest when a batch comparison RLA is conducted.
- 5) The manual count procedures detailed in .09(F) apply to batch comparison RLAs, but other RLA methods would utilize different procedures. We assume that the Board or State Administrator can promulgate emergency rules or guidance if it unexpectedly becomes necessary to use additional methods beyond batch comparison.
- 6) The language on “ballot control” in .09(D)(1)(a) would require local boards to store ballots in a container with only ballots from the same batch. While this provision is logical for storing election day and early voting ballots, it could be excessive for absentee ballots. Many U.S. jurisdictions scan and store absentee ballots in small batches of a uniform size to facilitate ballot accounting and audits. (While Maryland’s current voting equipment might not directly allow this practice, newer voting systems generally offer this feature.) This approach becomes impractical if each batch must be stored in a separate box. We would instead recommend allowing multiple batches of centrally-scanned ballots to be stored in a single ballot box, as long as each individual batch is clearly labeled and physically identified within the ballot box, and the batch label identifies all batches in the box. The definition of “Batch” in .01(B)(5) would also require revision. We would recommend removing “individually” from the definition so that it does not require batches to be “individually stored in a ballot box.”
A few much smaller points:
- 7) Parts of the regulations refer to “ballots,” “ballot cards,” or “paper ballots,” while other parts refer to “voter-verifiable paper records.” We believe that these terms are all intended to be synonymous. (We also noticed inconsistent hyphenation of “voter-verifiable” as well as “risk-limiting.” Per the statute, both would always be hyphenated.)
- 8) Under .09(C)(3) (“Requests by Local Boards to Audit Local Contests”), we suggest removing the requirement in (b)(i): “All Statewide contests must be completed prior to the commencement of the audit.” The purpose of this requirement is unclear, and (b)(ii) would guarantee that a local RLA does not impact the completion of a statewide RLA.
- 9) .09(D)(1) reads “as follow:” but should be “as follows:”.
- We suggest revising the language in .09(E)(1) so that it is consistent with the definitions in .01 (and to add a missing “the”): “Select a random seed to input into the risk-limiting audit software.”
- 10) .09(F)(1)(d) reads “Issue the teams batches or voter verifiable paper records[…],” but we believe it should read “Issue the teams batches of voter-verifiable paper records.”
Thank you for your consideration of our comments. We would be happy to discuss any of these suggestions with you.
Respectfully submitted,
Chrissa LaPorte
Deputy Policy & Strategy Director