FBI Turns to AI for Signature Checks on Seized Fulton County Ballots

The FBI is evaluating artificial-intelligence tools to compare signatures on roughly 116,000 mail-in ballot envelopes seized from Fulton County, Georgia, after a January raid that collected about 700 boxes of election material. Internal communications indicate the bureau plans to match envelope signatures with those on voter registration forms and, potentially, driver-license records. The effort is part of a broader redeployment of 260 analysts to the Fulton County probe, as outlined in a memo cited by *The New York Times* and MS Now. A White House spokesperson declined to comment, referring inquiries to the FBI, which also did not respond to a request for comment.

Propublica

Forensic specialists warn that AI-driven signature analysis is unproven at the scale required. Linton Mohammed, former president of the American Society of Questioned Document Examiners, testified that “signatures are one of the most difficult forensic sciences, and I don’t think AI is going to be able to do this.” He added that “signatures vary — unlike DNA or fingerprints.” Max Palmer, a Boston University professor, said there is “a high degree of noise” in the FBI’s material, questioning whether enough reliable data exist for accurate machine assessment. David Gerber, senior vice president of ParaScript, emphasized that technology alone is insufficient, noting that “signature matching in elections needs ‘the right technology with the right people managing the process… I don’t care how good the technology is.’”

Propublica

President Donald Trump has repeatedly claimed the 2020 election was stolen, tweeting in November 2020, “Must have signature check on envelopes now.” Conservative lawmakers responded with stricter verification laws, including Georgia’s Election Integrity Act of 2021. Activists’ allegations that Fulton officials “willfully” ignored signature checks helped prompt the FBI seizure, but an independent monitor and the States United Democracy Center have found those claims false.

Propublica

Sources within the FBI note internal pressure to produce fraud findings, yet staff developing the AI strategy warn that large-scale signature analysis lacks precedent and may be swayed by political influence. The FBI’s exact timeline for deploying AI, the specific commercial products under consideration (e.g., OpenAI or Anthropic), and the criteria for defining “evidence of fraud” remain undisclosed.

Propublica

• “ Ultimately, the tech specialist said, its results would turn on the threshold set for evidence of fraud: “It’s up to the builder of the system to define the guidelines.” — FBI tech specialist

Propublica

• “Must have signature check on envelopes now,” — President Donald Trump, Twitter (now X)

Propublica

• “Signatures are one of the most difficult forensic sciences, and I don’t think AI is going to be able to do this,” — Linton Mohammed, former ASQDE president

Propublica

• “Signatures vary — unlike DNA or fingerprints.” — Linton Mohammed

Propublica

• “There’s a high degree of noise” — Max Palmer, Boston University professor

Propublica

• “He said signature matching in elections needs “the right technology with the right people managing the process.” — David Gerber, senior vice president, ParaScript

Propublica