AI wars heating up ahead of early voting in SC

CHARLESTON, S.C. (WCSC) — Early voting starts Tuesday for state primaries, and if your phones and TVs weren’t filled with political messaging before, they’re about to be.

Artificial intelligence is changing the way political campaigns are run and the way voters view candidates. Like any new product, there’s good AI and bad AI; passive and active.

Generative AI is being used to create images and videos to trick voters through attack ads. In South Carolina, several prominent candidates have been the target of ads that depict made-up scenarios. For example, one that shows Attorney General Alan Wilson ignoring a phone call from President Donald Trump. Another ad shows Lt. Gov. Pamela Evette taking a lie detector test.

Alon Yamin, the CEO and co-founder at Copyleaks, an AI detection software, says things are starting to become more difficult to tell the difference between AI and human content.

“It’s becoming harder and harder to distinguish between AI and human content,” Yamin said. “The bar for fraud is just so low. Anyone can create whatever they want with AI at this point, and that creates a serious problem for the political landscape.”

There are a couple of things you can look for to help identify if a picture or a video is real. Check the details of a video. Smudging around features like hair and hands is a dead giveaway that a video is fake. If the lip sync is off, the video may be AI-generated as well. Be wary of low-quality videos. Blurry video is a good way to hide AI imperfections.

Yamin says AI detection software can help identify generative content in a way that isn’t detectable to the naked eye.

“Try to focus on the background and see if there are changes in the background as you would expect from a moving video. Things like that could be signs or tells,” Yamin said. “Unfortunately, unless you’re using an AI detection technology that could look at the frames and the pixels and tell you with a high level of accuracy if this is AI or not, you’re not able to really know.”

Lipsync is another big giveaway. If the words in a video don’t match the mouth they’re coming out of, you’re probably looking at a generated video.

Since AI is relatively new, the laws around regulating generative AI either don’t exist or aren’t keeping up with the advances that are happening at record speeds.

“We’re seeing different regulations in different states. Most regulations essentially require that individuals disclose when content is created by AI. Others have even forbidden completely the use of AI in a certain amount of time before the elections,” Yamin said. “I think really what’s missing is to have something on a federal level where all states are working in the same landscape.”

It’s the wild, wild west for AI and generative content—especially on social media. The best advice for those scrolling is to trust their instincts. If an image or video is unbelievable, it probably is.