ProPublica investigation reveals FBI explored using AI to review signatures on seized Fulton County 2020 mail-in ballots

On July 15, 2026, ProPublica reported that the FBI had explored using artificial intelligence to assess the validity of signatures on tens of thousands of mail-in ballot envelopes seized from Fulton County, Georgia -- the latest step in the Trump administration's reinvestigation of the 2020 presidential election. According to internal communications reviewed by ProPublica and an agency technology specialist familiar with the work, the effort focused on comparing signatures on ballot envelopes against signatures on other election documents, such as voter-registration forms. The tech specialist characterized the AI tooling as questionable and said any results would depend heavily on the threshold set for evidence of fraud. The signature-matching initiative was under discussion as recently as late June, though its current status was uncertain; a White House spokesperson referred questions to the FBI, which did not respond to ProPublica's request for comment. [1]

The exploration of AI signature review was part of a broader escalation of the Fulton County inquiry. In January, the FBI raided the county -- a Democratic stronghold -- and collected roughly 700 boxes of election materials, including about 150,000 mail-in ballots, of which some 116,000 were cast for Joe Biden, who defeated Donald Trump in Georgia by 11,779 votes. The bureau also mandated that 260 analysts be redirected from field offices nationwide to the Fulton County probe, and separately compared a spreadsheet of 175,000 voters against a commercial database to check, among other things, whether they were still alive. Signature matching as a fraud-detection method has been widely criticized: experts have raised concerns in litigation and in analyses of recent elections about how reliably it can distinguish a genuine signature from a forgery given natural variation in a person's handwriting. [1, 2]