AI-generated political ads are showing up on TV screens, experts say they are getting harder to spot

Posted

Artificial intelligence is turning up in political advertising this election cycle, including ads featuring images that are not real.

A recent report from The New York Times said $147 million has been spent to broadcast ads made with AI.

Ads airing this year have included an image of Senate candidate James Talarico reading a bible with AI generated audio, a fake video of Sen. John Cornyn on vacation and a video of former Senate candidate Jasmine Crockett and Cornyn dancing.

Al Tompkins, an author and journalist who teaches other journalists about AI, said artificial intelligence in political ads is becoming more common.

“There are a couple of reasons that AI has been very popular in this cycle. One is it’s cheaper,” Tompkins said. “Artificial intelligence allows you to have, for example, people saying things without having to go out and actually get people to say things. You’re able to show your opponent in the worst possible circumstances without actually having to have video of them.”

Tompkins said one problem is that AI gives people an excuse to discount real things they disagree with.

“‘I don’t like that, it must be AI’. When it wasn’t AI, they actually did say that, they actually did do that. That’s one of the costs of artificial intelligence is that we start only believing those things that we already believed,” he said.

Some ads disclose that they used AI, but Texas law does not require it.

One way to spot AI is garbled lettering, such as the text on a prison uniform and a corrections officer’s badge in one ad. Tompkins said AI is getting harder to spot as the technology improves.

He offered other tips for viewers.

  1. Does the video or image play on your emotions?
  2. Is there another place you can find this information? Is there a cited source?
  3. Does it make sense? Would a person have said or done this?
  4. Does it make a difference in your opinion, or how you’d vote?

Tompkins said attack ads target undecided voters with the goal of making people doubt what is real.

“We’re moving, I’m afraid, from skepticism to cynicism. Skepticism means I’m open to truth, but cynicism means no amount of truth will be enough. And I think that AI gives us cover for that,” he said.

Texas law says AI and deepfake ads cannot air within 30 days of an election. Tompkins said that could be challenged in court later.

More election coverage on KSAT: