TAMPA, Fla. — Gov. Ron DeSantis’ political campaign’s apparent use of fake, artificial intelligence-generated photos of former President Donald Trump hugging and kissing Anthony Fauci has raised concerns among experts about the new technology’s ability to deceive voters.
The pictures were part of a 40-second video posted last week by @DeSantisWarRoom on Twitter criticizing Trump — DeSantis’ chief rival for the GOP presidential nomination — for not firing Fauci, an infectious disease expert who was a part of the White House Coronavirus Task Force. Many Republicans say Fauci advised keeping businesses and schools closed unnecessarily throughout the pandemic.
The realistic-looking images were likely generated by AI technology and appeared to show Trump and Fauci embracing in various locations in the White House. The fake images were interspersed with real photographs of the two men together.
The video — posted by the Twitter account that does rapid response for the DeSantis campaign — was the latest example of the use in campaign attacks of generative artificial intelligence, which can create realistic images with a few prompts and clone human voices.
In May, Trump shared a two-minute parody video on Truth Social and Instagram of DeSantis’ presidential announcement that made it look like DeSantis was making his announcement alongside the devil, Adolf Hitler and others. And Donald Trump Jr. days later shared a video that replaced Steve Carell’s character from the hit show “The Office” with a figure that looked and sounded like DeSantis. In the scene, the characters of the sitcom mock Carell’s character for unknowingly wearing a woman’s suit.
Those videos appeared to have been created by other users, not the Trump campaign, and the altered images and audio appeared meant for the viewer to easily know they were doctored — although neither disclosed they were generated through AI.
The line between real and fake was blurrier with the fake photos of Trump kissing Fauci. And that could happen more and more in the future, said Janet Coats, director of the Consortium on Trust in Media and Technology at the University of Florida.
Coats said that generative AI will become more common, sophisticated and difficult to separate from real life as the technology improves and the 2024 election approaches.
Already, an array of free or cheap voice-cloning apps are available online and relatively easy to use. Many require a short audio sample of someone speaking, which can then be used to generate audio of someone saying whatever the user would like.
“We are going to be in an arms race until November 2024,” Coats said. “Right now, people have their toe in the water, but as they see what it can do, I think there’ll be more and more of it. The ability to check it is going to get harder, because we’re just going to be so inundated with it.”
The Republican National Committee sparked conversation about the use of AI in political messaging last month after posting a 30-second dystopian portrayal of what it said America would be like if President Joe Biden is reelected. The video contains a disclaimer in the top left corner that the images were generated by AI.
The American Association of Political Consultants in May condemned the use of deceptive AI-generated content in political campaigns.
After the DeSantis campaign posted the video of Trump and Fauci, a Twitter community note flagged the post, warning readers that the video contained fake, AI-generated images. AFP Fact Check was the first to flag the fake photos.
Some Republicans pushed back against the DeSantis campaign.
“Smearing Donald Trump with fake AI images is completely unacceptable,” said Sen. J.D. Vance of Ohio, who has endorsed Trump for another term as president. “I’m not sharing them, but we’re in a new era. Be even more skeptical of what you see on the internet.”
U.S. Rep Marjorie Taylor Greene of Georgia, an avid Trump supporter, also criticized the ad, saying the post should be taken down.
There was also speculation that a May 24 DeSantis ad used AI to include fighter jets flying over the governor during a speech. It’s unclear whether the added jets were the product of AI or manual digital alteration.
On June 8, seemingly pushing back against criticism, DeSantis rapid response director Christina Pushaw posted a screenshot of a Trump post on Truth Social featuring Ron DeSantis riding a rhino in what appeared to be an insinuation that DeSantis is a Republican in name only, or RINO. Pushaw wrote above the screenshot: “I think this might be an AI-generated image. Who knows?”
Asked about the Fauci-Trump images, a Trump spokesperson responded, “Fake images from a fake campaign for a fake candidate. This stunt had the ‘please clap’ energy of Ron DeSanctimonious’ mentor, Jeb Bush.”
The DeSantis campaign did not respond to multiple emails requesting comment.
On June 5, Democratic U.S. Rep. Ritchie Torres of New York introduced a bill that would require a disclaimer for content generated by AI.
Although the public knows that AI-generated images and audio exist, the risk is that, as the technology improves and becomes more common, such content could become “visual clutter” that people don’t take in or look at critically, Coats said.
Campaigns may also try to use AI to backtrack on previously stated positions by altering real content, or by claiming that the real content was AI-generated, Coats said.
“If there’s so much of it that you can no longer discern easily that this is not real, this person did not say this, this is misinformation, disinformation ... it gets really, really complicated,” Coats said.
