The FBI explored using artificial intelligence technology to analyze signatures on mail-in ballots seized during investigations related to the 2020 election, according to reporting by ProPublica. The agency considered deploying AI systems to review ballot signatures, a practice that raises substantial concerns about accuracy and reliability in election administration.
The use of AI for signature verification presents technical and legal hazards. Signature recognition algorithms frequently produce false matches and false negatives, particularly across demographic groups. Courts have repeatedly expressed skepticism about AI evidence absent rigorous validation studies and transparent methodology. No peer-reviewed research establishes that commercial AI signature-verification systems meet forensic standards required for evidentiary use.
Federal law governing election disputes contains strict procedural requirements. Under 52 U.S.C. Section 20511, ballots cannot be rejected based on signature mismatches without proper notice and opportunity for voters to cure. The Help America Vote Act similarly mandates specific processes for ballot challenges. Using unvalidated AI systems to flag ballots for rejection could violate these statutory protections and constitutional due process guarantees.
The FBI's consideration of this technology implicates broader election integrity concerns. Mail-in ballot signature verification already struggles with consistency across jurisdictions. Introducing algorithmic analysis without transparent standards or external auditing compounds existing vulnerabilities rather than solving them.
The implications extend beyond 2020 disputes. As election administration increasingly incorporates technology, agencies must establish clear validation standards before deployment. The National Academy of Sciences, the Election Assistance Commission, and voting security experts consistently warn against adopting election technologies without rigorous testing and public scrutiny.
This development underscores tensions between efficiency and accuracy in election administration. While technology can streamline processes, premature or inadequately tested deployment carries risks to ballot integrity and public confidence. Federal agencies considering such tools should follow established protocols for evidence evaluation and seek independent technical review before operational use.