Washington grapples with AI deepfakes on the campaign trail

The 2024 campaign cycle is expected to unleash a torrent of AI-generated “deepfake” videos and images capable of deceiving voters — but the early attempts in Washington to stem the problem have already hit political snags.

Election watchdogs hope to clear one such hurdle Thursday. The Federal Election Commission is scheduled to vote on a petition, initiated by advocacy group Public Citizen, that calls for banning political campaigns from distributing fake audio, video and images of their opponents.

Public Citizen notes that deepfakes have already appeared in ads supporting Florida Gov. Ron DeSantis, including fake photos that depict former President Donald Trump embracing Anthony Fauci. Such forgeries have alarmed election officials and advocates alike for their potential to confuse voters and skew election outcomes.

“They're already good enough to fool you if you're not paying careful attention,” Public Citizen President Robert Weissman said in an interview on Thursday’s episode of the POLITICO Tech podcast.

“We're going to have very high-quality deepfakes in advance of 2024,” he added. “And in the absence of some federal prohibition on the use of deepfakes, I think it's a near certainty that political actors of all political stripes will use them.”

Public Citizen is calling on the FEC to specify that existing rules against “fraudulent misrepresentation” apply to fake video, audio and images, which due to advances in artificial intelligence have become increasingly realistic and easy to produce.

Thursday’s expected vote is a preliminary step that would determine whether the FEC seeks public comment on the proposed ban. The so-called notice of availability does not commit the commission to taking further action and such moves generally face little opposition.

But an earlier version of the petition failed in June after a similar vote split along party lines, with the FEC’s three Republican commissioners arguing that Public Citizen did not properly cite the regulation in question and, more critically, that the requested rule change would exceed the agency’s authority.

“Congress has given us very sharply limited authority,” Republican Commissioner Allen Dickerson said at the June meeting. “Our jurisdiction on this point is limited to instances where a campaign broadly misrepresents itself as acting on behalf of any other candidate or political party.”

“Instead of coming to us, you take this up with Congress,” he continued. “I wish it luck.”

But in Congress, the issue has also taken on a partisan tone. Democratic Rep. Yvette Clarke (D-N.Y.) introduced a bill in May that would mandate political ads disclose the use of AI-generated material. Then in July, 50 Democrats signed a letter to the FEC backing Public Citizen’s petition.

“The biggest hurdle preventing Washington from taking action at the moment is who owns the problem,” said Katie Harbath, the senior adviser for technology and democracy at the International Republican Institute.

“The FEC kicked it back to Congress. Congress doesn’t seem to be in a huge hurry. The Biden administration says they want to use existing laws to regulate it. So there’s a lot to sort out there,” continued Harbath, a former public policy director at Facebook.

Harbath anticipates mainstream media will vigilantly fact check AI-generated deepfakes and that the technology used to identify spoofs will evolve along with the technology used to make them. Even still, the “right systems” are needed “to prepare for the things you can imagine might happen and how to react to the things you didn’t imagine,” she said.

“There’s no real deadline or no turn-back-now sign as these problems will last well past the election, but that doesn’t mean we can wait,” Harbath said.

Weissman said an updated petition that Public Citizen submitted last month should address Republican commissioners’ concerns and he is “cautiously optimistic” they will at least allow a period for public comment.

Still, the vote in June suggests Republican commissioners are hesitant to entertain new rules on artificial intelligence. And even if they collect comments, partisan gridlock has halted the FEC from undertaking any serious rulemaking in recent years.

But, Weissman contends, the problems posed by deepfakes are not limited to one party.

“If there are Republicans or if there are Democrats who think that only Republicans are going to use this tool, they are wrong,” Weissman said. “It is a near certainty that all political actors, all the outside consultants, all the outside organizations, will use this tool if it's legal to do so.”

And he predicts unfettered use of deepfakes will cause people to doubt all media, legitimate or otherwise, leading them to simply decide for themselves what is real or not based on whether it underpins their existing political views.

“The deeper, more fundamental problem is that people stop believing what's before their eyes,” Weissman said on POLITICO Tech. “Every bad tendency we have for tribalism will be reinforced.”

Annie Rees contributed to this report. 

Subscribe to our new POLITICO Tech podcast for a daily download on the ways technology is disrupting politics and policy.