INDIANAPOLIS (WISH) — Purdue University researchers told I-Team 8 they're already seeing an increase in the amount of deepfake content on the web as the 2024 presidential election heats up.
“Disinformation and misinformation are as old as politics itself. We have the power here to proliferate increasingly credible content in ways we haven't been able to before,” said Daniel Schiff, a professor at Purdue University.
Schiff, whose research focuses on the policy and ethics of artificial intelligence, has tracked deepfakes online and spoke to I-Team 8 about some of the tools being used to create fake content.
- An automatic voice generator that converts user input into audio.
- Image generator.
- A generator to create fake person face photos.
“So you can take one of these images, or you can take 1,000 of them, and create a social media profile that shows whether you support or oppose a particular political candidate,” Schiff said as he pressed the button multiple times, creating a new mugshot with each click.
Schiff's content database contains over 400 pieces of content. Some of the content is completely fake, some is real but is said to be fake, and some has been altered to look different from the original. The content takes a political stance.
“In a political context, the biggest concern is that deepfakes could be used to influence election outcomes – to misinform people about political positions and the electoral process, and to incite animosity between social and political groups,” Schiff said.
Who is the villain in this situation?
“The typical assumption here would be a hostile foreign power, and indeed there is a lot of activity from foreign parties trying to influence elections around the world, but the vast majority of misinformation and disinformation comes from within the country — from political parties, campaigns, campaign proxies or independent groups that have a vested interest in the fight,” the assistant professor of political science said.
Schiff's wife and research partner, Kaylin Jackson Schiff, also an assistant professor of political science at Purdue University, told I-Team 8 that the Indiana Legislature has been working on a bill to restrict AI-generated deepfakes during election time, but they were unable to finalize a bill during the last legislative session and it is unlikely to go into effect in November.
“In many states, the legislative session is short, maybe two or three months,” Jackson Schiff said. “So what happens if technology advances in the next few months? You have to wait until the next legislative session.”
In the meantime, it is everyone’s responsibility to critically analyze anything online and determine whether it is fake content or not.
“Any time you receive information, especially if it's shocking or inflammatory or if it confirms your suspicions — if it's about a politician you don't like and it criticizes that politician — you should be cautious. It may be inaccurate,” Schiff said.
To avoid being fooled by deepfake content, the researchers suggest visiting at least two other trusted news sources to verify that it's true.
