The FBI has explored using artificial intelligence to check the legitimacy of the names on tens of thousands of mail-in ballot envelopes taken from Fulton County, Georgia. This is the most recent step in the Trump administration’s unprecedented push to promote the 2020 election.
According to internal contacts reviewed by ProPublica and a work-trained company technical expert with experience, the effort focuses on comparing signatures on registration forms and other election documents. Donald Trump has much asserted that the 2020 presidential election was a fraud to him without any proof.
He has consistently asserted that Georgia was a voter fraud state, losing to Joe Biden by only 11, 779 votes. About 150, 000 mail-in votes were collected in Fulton County, a Democrat redoubt, in January, out of which 116, 000 went for Biden. Trump is scheduled to address the Fulton County research in a speech on Thursday about voting machine vulnerabilities and federal election security.
Later in June, the signature-matching program was being discussed, but its current status is a mystery. A White House spokesman referred the inquiries to the FBI instead of answering them from ProPublica. The FBI did not respond to a request for comment.
According to an organization note that ProPublica reviewed, the FBI has mandated that 260 researchers be redirected from industry practices statewide to focus on the Fulton County sensor. The letter was recently reported by MS Now and The New York Times. According to a report from MS Then, some FBI researchers were let go rather than put to work on the project. Their research includes, among other things, comparing a worksheet of 175,000 voters to a professional database in order to determine whether they are still intact.
The precision of personal matching, whether done by people or technologies, is still a contentious issue.
After analyzing recent election results, experts have raised serious concerns about how exactly signature matching may detect voter fraud by examining the size and slant of letters made under various circumstances.
The FBI technologies professional told ProPublica that the commission has the ability to assess images and that a signature-matching analysis may be” somewhat accurate” if it begins with a large enough data.
In the end, the tech expert claimed that the results do “turn on the threshold for evidence of fraud”:” It’s up to the system’s builder to determine the guidelines.”
Some FBI agents who are developing that strategy are attempting to avert the political pressure to implicate fraud in Fulton County by highlighting the limitations of large personal research, arguing that despite being used in adult voter fraud investigations, they haven’t been done on this scale, according to the source. However, agency leaders have kept pushing.
The FBI has a lot of concerns about how the examination’s results will reflect political influence, building on previous attempts by the administration to remove barriers intended to prevent the federal government from influencing elections. According to the source, “everyone is of the opinion that they are going to continue to pursue proof of fraud, regardless of whether they find anything.”
When more Democrats used mail-in ballots than Republicans during the 2020 election and COVID-19 pandemic, there was controversy surrounding signature matching. Trump allegedly made up false allegations that Georgia officials facilitated fraud by allowing signature matching on mail-in ballots. He wrote in late November 2020 on the social media platform Twitter, known as X, that “must have signature check on envelopes now.” Far more votes than necessary for the flip of the election to him.
Then, in response, conservative lawmakers advocated stringent signature-matching laws throughout the country, including Georgia’s Election Integrity Act of 2021, which they justified by citing “many electors concerned about allegations of rampant voter fraud,” as well as” subjective signature-matching requirements.” The bill replaced signature matching with more stringent forms of verification, such as requiring mail-in voters to provide their driver’s license number or copies of licenses.
Officials typically have only seconds to determine whether a signature on a mail-in ballot envelope is the same as it appears on other government records, sometimes relying solely on a letter’s shape or proportions, despite jurisdictional differences in signature-matching procedures. According to research, signatures can change over time as people get older or have illnesses like a stroke, and signatures made in different settings can differ, such as when they were carefully executed in government offices or quickly delivered in a mail-in ballot envelope at home.
According to research conducted by journalism organizations, signature matching results in disproportionately high voter turnout rates for people of color, as well as those who are new, young, old, politically unaffiliated, or disabled, for a variety of reasons. More legitimate ballots may occasionally be disqualified due to signature-matching efforts. A political scientist who testified as an expert witness in a lawsuit challenging an Ohio signature-matching law in 2020 said his analysis suggested that 32 legitimate ballots were blocked for each illegitimate one.
” I don’t think AI will be able to do this,” said Linton Mohammed, a former president of the American Society of Questioned Document Examiners, a professional association for forensic document examiners.” Signatures are one of the most difficult forensic sciences, and I don’t think AI will be able to do this. ” Signatures vary, in contrast to DNA or fingerprints.”
The National Academy of Sciences ‘ 2009 report on the state of forensic sciences stated that” the scientific justification for handwriting comparisons needs to be strengthened.
The use of computer analysis has increased in sophistication and provides a necessary and practical check against fraud, claim experts who support signature matching. However, they do point out that the technique only works as well as the people who employ it.
According to David Gerber, senior vice president at ParaScript, a business that sells signature-matching technology to banks and a number of other organizations engaged in elections, “you can make your head explode.” He does, however, insist that trained human specialists be able to check the machines by reviewing about 10 % of the cases. He claimed that the “right technology with the right people managing the process” is required for signature matching in elections. Both of those pieces are equally significant. I don’t care how good the technology is if bad people are running a bad process.
According to most experts consulted by ProPublica and the FBI tech specialist, using AI to check signatures on ballot envelopes is a relatively new frontier. The bureau, according to the tech specialist, had discussed how to work with experts from all levels of government and had considered whether to use commercial products like those made available through OpenAI or Anthropic. According to the source,” this is a new and novel approach” for the bureau, one reason being” they are shopping it around.”
Nonspecialized AI software was unlikely to be able to perform this work, according to Gerber, Mohammed, and other experts on signature-matching technology.
Experts claimed that while an AI system might have some advantages over humans, its accuracy would ultimately depend on the quantity and quality of the comparisons ‘ signatures. The FBI’s analysis would only compare the signature on a voter’s registration form to the signature on the ballot envelope, according to internal communications reviewed by ProPublica. This would, according to experts, significantly increase the likelihood of discrepancies. The source noted, however, additional records, such as driver’s license signatures, could also be incorporated into the analysis. For industries like banking, certified specialists and high-concept computer systems typically compare numerous signatures.
According to Max Palmer, a professor at Boston University who has studied mail-in ballot signature matching,” there’s a high degree of noise” in the materials the FBI has. ” I’m not sure there’s enough data or signal to do better,” he said.
Conservative activists have long advocated for more thorough examination of Fulton County’s 2020 mail-in ballots, alleging in a report that local election officials “willfully” followed signature verification procedures. According to ProPublica, their claims played a significant role in the FBI’s ballot seizure in January.
The claims made by the activists about signature matching and other irregularities were ruled out by an independent monitor and the States United Democracy Center, a nonprofit organization working to protect the integrity of elections.
The first article on ProPublica titled The FBI Has Looked at Using Questionable AI Tech to Review Signatures on Seized Mail-In Ballots.
