AI could expose how Georgia voters cast their ballot, researchers warn

Max Springer, a Princeton researcher, discovered that publicly available election records and AI could potentially link voters to their ballots in Georgia. Springer used a $20 AI subscription and data obtained from an Open Records Act request to create a pipeline that could analyze and identify secret ballots across the state. The issue has raised concerns among Georgia election officials as early voting begins for the crucial midterm election.

Ballot secrecy is crucial for protecting voters' choices from potential influence or intimidation. However, Springer's test demonstrated that AI could be used to uncover who people voted for, using Georgia's public election records. Georgia's State Elections Board held an emergency meeting to discuss how AI had made it easier to identify voters through their ballot code.

The flaw lies in the way the machines record ballots, which are not random enough and can be matched with other information like sign-in data or surveillance footage. Election officials have been requesting funding to upgrade Georgia's voting system for years, but their requests have gone unanswered due to a political feud. Despite the strong constitutional requirement for secret ballots, Georgia's law also demands transparency for election records, creating a conflict.

Some election workers are considering measures like shuffling printed ballots to prevent the problem, but the issue remains unresolved.