As Florida voters go to the polls on Tuesday and later cast votes in November for the presidential election, a factor they may need to contend with are political ads that may contain deceptive artificial intelligence, or “AI.”
In fact, it’s already happening.
Last summer, the federal super PAC created for Gov. Ron DeSantis — Never Back Down — manipulated the voice of Donald Trump to attack the former president in one such video ad that aired in Iowa, without suffering any consequences. (DeSantis has since suspended his presidential campaign.)
That’s why Florida is among the dozens of states that are currently considering laws to provide voters more information when AI-generated content makes its way into political ads. Indiana became the latest to regulate so-called “deepfakes” last week, joining New Mexico, Texas, Minnesota, Michigan, California and Washington, according to Public Citizen, a consumer advocacy organization based in Washington, D.C.
As a way to give warning to Floridians who may encounter such ads, state lawmakers in Tallahassee recently approved a bill (HB 919) that requires a disclaimer to be included in any political advertisement that uses generative artificial intelligence.
DeSantis will need to approve, or disapprove, the measure.
Language too weak
Under the legislation approved by the House and Senate, the disclosure must include this statement: “Created in whole or in part with the use of generative artificial intelligence (AI).”
But critics say that language is far too weak.
“It doesn’t say that the information is false,” says Amy Keith, the executive director of Common Cause Florida, about the disclaimer. “It doesn’t say that it’s been manipulated. It just says that the advertisement has been created with the use of generative artificial intelligence, and with so much stuff being created with the support of artificial intelligence by now that we feel that doesn’t really inform Floridians of what they’re looking at.”
The bill lists the disclaimer requirements for each specific medium, according to a bill analysis:
- For TV or video ads: The disclaimer must be clearly readable throughout the communication and occupy at least 4 percent of the vertical picture height.
- For internet public communications that include text or graphic components: The disclaimer must be viewable without the user taking any action and be large enough to be clearly readable.
- For printed communications: The disclaimer must be slated in bold font with a size of at least 12 points.
- For graphic communications: The disclaimer must be large enough to be clearly readable but no less than 4 percent of the vertical height of the communication.
If a disclaimer is not listed, a complaint would need to be filed with the Florida Election Commission. If the commission found the complaint to be valid, it would then be referred to a state attorney’s office for potential prosecution.
Miami-Dade County Democratic Rep. Ashley Gantt was one of the eight Democratic House members during the legislative session who voted the AI bill down. She said the language was confusing in terms of who would be held liable for any sanctions. “The bill itself wasn’t really clear about who will bear the responsibility,” she told the Phoenix last week.
The bill says “[i]n addition to any civil penalties provided by law,” the person identified as having paid for, sponsored, or approved the AI-generated ad who fails to place the required disclaimer is also subject to a first-degree misdemeanor. And it says that a person identified as “paying for, sponsoring, or approving” the ad in question would be held responsible.
Orlando-area Democratic Rep. Anna Eskamani voted for the bill on the House floor, but she says that in Florida’s current election environment, it may be extremely difficult to determine who is responsible for running an AI-generated ad with, or without a disclaimer.
“Let’s just say that a random PC (political committee) uses AI and they mail a flier to 30,000 doors,” she says. “It didn’t have a disclaimer on it. How do you even find out what’s behind that because some of these PCs hide behind layers and layers of just cloudiness. It might be a random person who lives in another state. It’s just kind of intentionally an opaque system that allows for unethical decision making and it’s not like a disclaimer requirement is going to stop that from happening.”
Deepfake ads will be rampant
Common Cause’ Keith says another problem with the bill is that there is no process to take down the ad as soon as possible, meaning it could continue to be broadcast without a disclaimer.
“Say you’re a candidate and you see a manipulated ad that is manipulating what you yourself said or did and it’s showing something false about what you did,” she says. “So you’re having more Floridians exposed to that false information while it works its way through the process.”
Ilana Beller is the organizing manager for the democracy team at Public Citizen. She notes that absent any federal legislation on the issue, it’s incumbent for states to pass laws requiring some type of disclosure for political ads that contain generative artificial intelligence, or she says, such deepfake ads “will be rampant.”
“I think that it’s a deterrent to bad actors that they now know that a) there are some penalties that they will have to pay if there isn’t a disclosure and b) they’ll have to circulate them with a disclosure, so they’re basically telling on themselves if they’re following the law. Saying ‘hey, this is manipulated, fraudulent content,’ that will be a huge deterrent for the vast majority of people who want to put these out there who want to convince people that they’re real.”
One expert told the Phoenix that it’s simply too early to determine if these disclaimers will work.
“We have very little direct, empirical research looking at the effectiveness or the impact of these sorts of disclosures in generated AI in political ads,” says Scott Babwah Brennan, the head of Online Expression Policy for the Center on Technology Policy at UNC-Chapel Hill.
“We don’t know if they work, we don’t know if there are other sorts of impacts that they might have, let alone in getting to the level of ‘how should they be worded?’ And the differences that different language might have. That’s not to say that they’re ineffective or there aren’t ways to make them or less effective, but rather we just don’t know,” he told the Phoenix.
While there is considerable concern that deepfakes could be a part of the presidential election this year, Brennan believes they might have a bigger impact on local elections.
“In smaller down-ballot races there’s less advertising, less attention, less oversight,” he says. “There’s just less communication about the candidates.”
In the case of the DeSantis super PAC ad that manipulated the voice of Donald Trump to make it sound like he was criticizing Iowa’s governor Kim Reynolds (who had endorsed DeSantis), a spokesperson for the super PAC told PolitiFact that they had “utilized technology to give voice to Donald Trump’s words” from a Truth Social post. That prompted the Trump campaign to say that it was “a desperate attempt” to deceive the public.
The League of Women Voters of Florida supports the legislation.
The post As the election year ramps up in Florida, concerns grow about AI in political ads appeared first on Florida Phoenix.
This article originally appeared here and was republished with permission.
