Source: Fortune

Security researchers warn that advances in artificial intelligence make it more urgent to address a software flaw in Georgia’s touchscreen-and-scanner voting system that can allow electronic ballot images and cast‑vote records to be put back into the order they were cast, enabling linkage to voter check‑in and early voting lists obtained via public records. The vulnerability does not change votes or election totals but can reveal how specific individuals voted, posing a risk to the state constitutional guarantee of secret ballots and creating potential openings for legal challenges.

Princeton researcher Max Springer demonstrated that a publicly available AI assistant could reverse the randomization for Georgia’s May primary and identify additional records needed to match names to ballots, producing “buckets of ballots and potential voters.” Voting‑technology specialist J. Alex Halderman and Verified Voting’s Mark Lindeman cautioned that AI automates the technical work and that Georgia’s statewide use of the system plus open‑records access creates a “perfect storm.” Georgia officials say they have taken steps—directing certain open‑records requests to the secretary of state for redaction, working with a vendor to scramble the order of public files, and limiting releases—but advocates like Marilyn Marks argue that withholding records is not sufficient and that the proper fix is to stop recording identifiers or change scanning procedures. A State Election Board proposal to require the vendor patch was rejected in August amid concerns about certification, timeline, authority, and funding; board members also noted felony penalties for revealing how someone voted as a deterrent. With elections approaching, researchers and advocates urge either technical patches, changed procedures (for example central shuffling and rescanning), or other measures to prevent AI‑enabled de‑anonymization of ballots.