Georgia takes on deepfakes that could trick voters in ‘24 election

Tech impersonating candidates would be criminalized in state
Rep. Brad Thomas, R-Holly Springs, talks about deepfakes — computer-generated versions of politicians used to spread misinformation —  during a House Technology and Infrastructure Innovation Committee meeting earlier this month. (Natrice Miller / Natrice.miller@ajc.com)

Credit: Natrice Miller/AJC

Credit: Natrice Miller/AJC

Rep. Brad Thomas, R-Holly Springs, talks about deepfakes — computer-generated versions of politicians used to spread misinformation — during a House Technology and Infrastructure Innovation Committee meeting earlier this month. (Natrice Miller / Natrice.miller@ajc.com)

Fearing that voters could be tricked by misinformation from computer-generated versions of politicians such as Joe Biden or Donald Trump, Georgia lawmakers are rushing to stop malicious spoofs in time for the 2024 presidential election.

Legislation to criminalize deepfakes, such as a robocall that mimicked Biden’s voice to discourage people from voting last month in the New Hampshire primary, targets this latest risk of election subversion.

“When you take someone’s image and likeness, the way they speak and move, having them do something they never did, that’s fraud to me,” said state Rep. Brad Thomas, the Republican sponsor of the bill. “It’s scary. This bill provides the guardrails that people need and election integrity needs.”

While deepfakes haven’t infected campaigns in Georgia yet, legislators have seen their recent spread: the Biden robocall, a faked explicit image of Taylor Swift, and manipulated audio of party leaders in Slovakia conspiring to rig an election. None of them was authentic.

With the support of House Speaker Jon Burns, the Georgia deepfake bill is scheduled for a House vote Thursday.

“These proposed measures will be a great first step toward addressing an extremely important emerging technology and ensure artificial intelligence will be a net positive for Georgia families, businesses and communities,” said Burns, a Republican from Newington.

Credit: Natrice Miller/AJC

icon to expand image

Credit: Natrice Miller/AJC

The danger of deepfakes — enabled by increasingly common artificial intelligence tools — jeopardizes voters’ faith in elections and candidates, using speech as a way to try to tamper with voting behavior and maybe even results, said Thomas of Holly Springs.

“The risk is real. We need to get to a point where citizens are not derailed by what is essentially fancy spam,” said Josh Lawson, director of artificial intelligence and democracy at the Aspen Institute, a think tank focused on U.S. and global issues. “That’s going to require a whole-of-society effort to regulate what we can, penalize bad acts and create a social expectation to verify information and use trustworthy sources.”

Five states have already enacted deepfake election laws, and Thomas said he hopes Georgia’s bill can be a model for the nation.

Under House Bill 986, it would become a felony to broadcast or publish deceptive information within 90 days of an election with the intent to influence a candidate’s chance of being elected, create confusion about election administration or influence the result.

Satire, parody and legitimate journalism would be exempted. Campaign ads would be required to include disclosures if they include AI-generated content. Violations would come with a sentence of two to five years in prison and a maximum fine of $50,000, and the bill would allow a judge to bar individuals from distributing deceptive media.

Credit: Natrice Miller/AJC

icon to expand image

Credit: Natrice Miller/AJC

A new deepfake law must balance free speech rights with the need to prevent election interference, said Sarah Hunt-Blackwell of the American Civil Liberties Union of Georgia.

“We’re really in new territory here,” Hunt-Blackwell said. “The Founding Fathers of our country, when drafting our Constitution, weren’t thinking that technology could re-create people’s image and voice of something that didn’t happen in real life.”

Several major technology companies — including Google, Meta and Microsoft — signed a voluntary agreement last week to adopt “reasonable precautions” to prevent artificial intelligence tools from disrupting elections. The agreement was largely symbolic and lacked any binding requirements on the companies.

Georgia lawmakers’ efforts to combat election misinformation are advancing through the General Assembly, along with several other bills that would change the state’s voting laws, including proposals to increase the number of election audits, eliminate computer codes from ballots and add watermarks to ballots.

“In the old days, you could tell when something was fake,” said Robert Smith, general counsel for the Prosecuting Attorneys’ Council of Georgia, which has testified on the bill but hasn’t taken a position. “The technology is so good these days. It’s out there. My eyes have been opened based on this bill.”

After the faked Biden robocall in the New Hampshire primary, the Federal Communications Commission ruled that calls made with AI-generated voices that target consumers are illegal.

Thirty-nine states have introduced or passed legislation to address deepfakes, said Jonah Minkoff-Zern of Public Citizen, a liberal consumer advocacy organization.

“Deepfakes and misinformation are important threats to democracy,” Minkoff-Zern said. “The ability to have someone do it sitting in their living room is getting easier and easier, so we’re glad to see states taking this up across the country.”