The NRSC made a Texas Democrat say things he never said. The disclaimer fits in a corner.

A political ad from the National Republican Senatorial Committee opens with Texas State Representative James Talarico appearing to stand in front of a Texas flag, beaming. "Radicalized white men are the greatest domestic terrorist threat in our country," the U.S. Senate candidate seems to say. Talarico never filmed that video. The NRSC generated it with AI. The words "AI generated" appear in easy-to-miss font in the lower right corner.

This is where campaign ads are now. Republicans appear to be using deepfakes more than Democrats this cycle. The NRSC has released at least three such ads. A Republican committee in Loudoun County, Virginia released AI-generated videos attacking Democratic Governor Abigail Spanberger. The Collins campaign in Georgia created a deepfake of Democratic Senator Jon Ossoff appearing to say he'd never seen a farm except on Instagram. On the Democratic side, California Governor Gavin Newsom has used AI-generated videos to mock Trump — but the national Democratic campaign committees have not yet moved the same way on midterm ads.

There is no federal regulation governing AI in political messaging. Twenty-eight states have passed laws, mostly focused on disclosure rather than outright bans. Research suggests those disclosures don't work — voters still get influenced even after being told a video is AI-generated. "I think that the types of damage that we can do to the rigor and credibility of elections... very much risks being supercharged," said Daniel Schiff, a Purdue University professor who has studied thousands of deepfakes.

The NRSC said Democrats were "panicking after seeing and hearing James Talarico's own words." Talarico's campaign noted that his opponents "spend their time making deepfake videos to mislead Texans." He did not, as far as anyone can tell, film that video.