State Rep. Nick Pisciottano wants to make sure deepfake videos don’t interfere with Pennsylvania’s elections in the future.
The Allegheny County Democrat has introduced legislation that would make the dissemination of political deepfakes within 90 days of an election subject to civil liability and potential fines.
Deepfakes are videos and photos that use a form of artificial intelligence called deep learning to create convincing fake likenesses of people and events.
According to Sensity AI, an artificial intelligence company, the number of deepfake videos on the internet is growing exponentially, according to a report released this month.
But how much of a problem are deepfakes when it comes to Pennsylvania politics? And is an outright ban of deepfake dissemination the right way to curb election-related disinformation?
According to Pisciottano, they’re a big one, and must be confronted.
“They create a distrust in government and in our institutions that is not easily or quickly restored,” he wrote in a memo seeking co-sponsors for his proposal.
And as “technology advances and artificial intelligence-driven video editing software becomes more widespread, misinformation tactics will evolve and the use of so-called ‘deepfakes’ will insert themselves into our daily lives,” he continued.
Pisciottano conceded that these videos have yet to have a measurable impact on Pennsylvania’s elections. But as deepfakes become more and more widespread — according to artificial intelligence company Sensity AI, the number of deepfake videos on the internet is doubling every six months — he fears they will.
Deepfake videos can have troubling consequences when shared widely online. They have notably been used for harassment and intimidation, such as creating explicit or incriminating photos of celebrities as well as normal folks — most often women.
In Pennsylvania, a Bucks County woman was charged with cyber-harassment for allegedly creating deepfake images of teen girls on her daughter’s cheerleader squad, depicting them naked, smoking and drinking, the Philadelphia Inquirer reported in March.
Deepfakes have touched the world of American politics too, as several videos deepfake videos of politicians have gone viral online. A notable example is a doctored video of U.S. House Speaker Nancy Pelosi, D-Calif., appearing drunk and slurring her words that went viral on right-wing internet circles.
It’s an issue that Congress has taken note of as well. The National Defense Authorization Act of 2021 orders the Department of Homeland Security and the Pentagon to study the technology behind and potential harms caused by deepfakes.
Pisciottano fears that deepfake videos could be used to influence the outcome of an election.
A particular problem, he says, is the possibility of a viral deepfake in the waning days of a campaign, where candidates have little time to refute a fake video of them doing or saying something incriminating before voters hit the polls. Though such a scenario has not yet come to pass, Pisciottano wants to act “proactively rather than reactively” to prevent the spread of misleading deepfakes.
“I think that these deepfake videos are going to be problems for us down the road,” Pisciottano said.
“Both sides of the aisle are very politically polarized and open to accepting information that they already believe to be true,” he added. “Deepfakes make that worse.”
Technology platforms are attempting to keep up with deepfake technology by designing tools that can detect deepfakes — but it’ll be a race between the two technologies and a question of whether detection software can keep up.
“The real problem is we don’t know which ones are deep fakes and which ones are not,” Shyam Sundar, a media professor and disinformation expert at Penn State University, told the Capital-Star. “It’s a very real problem especially given there’s so much polarized partisan politicking that goes on especially around election time.”
Sundar’s research at Penn State has found that faked videos have more of a capacity to deceive the public than text-based misinformation. In other words, seeing is believing, and faked videos have a startling capacity to deceive the public.
“A video fake is much more believable than fake audio or fake text,” Sundar said. “It’s right in front of their eyes. Your eyes cannot lie.”
But Pisciottano’s bill, though it aims to curb misinformation, runs the risk of coming into conflict with First Amendment protections, according to Sundar.
A similar law in California has faced questions about its enforceability because many forms of political speech are protected by free speech, and deepfake political videos may well be considered protected free speech.
Pisciottano said he modeled the bill’s language off of existing laws that limit revenge porn and defamation to avoid legal challenges to the bill. The 90-day limit on deepfake circulation also aims to target election misinformation specifically without threatening other forms of free speech.
Still, the bill could face legal challenges if it passes in its current form.
“I don’t think it will stand the test of the First Amendment,” Sundar said.
Sundar thinks a more effective way to combat deepfake misinformation is through media literacy campaigns that encourage people to question the source of their information.
Pisciottano said that he is open to changes to the bill and is aware of the potential First Amendment questions.
“[The bill] could certainly change depending on feedback from everyone,” Pisciottano said, adding that “there’s an acceptance on all sides of the political aisle that [deepfakes] could potentially be a problem.”
Our stories may be republished online or in print under Creative Commons license CC BY-NC-ND 4.0. We ask that you edit only for style or to shorten, provide proper attribution and link to our web site. Please see our republishing guidelines for use of photos and graphics.