AI deepfakes don’t have to be flawless. They have to look believable enough to make you react. We spoke with CNET reporter Katelyn Chedraoui about where deepfakes are now—and what gives them so much potential as vehicles for disinformation. Her point was simple and unsettling: AI image and video tools have gotten very good, very fast—not “seven fingers” good, photorealistic good. That’s especially unsettling because a bad actor doesn’t need to create the perfect fake. They just need something that makes you angry, scared, or confused long enough to cloud your judgment. That can be a video; a robocall; a text about a polling place changing. Sometimes the most effective deception is also the most mundane. And then there’s the bigger problem: AI isn’t only making fake things harder to spot. It’s also making real things easier to doubt. Co-hosts Jim Ludes and Mark Jacobson also discuss the recent U.S. indictment of Russian intelligence agents accused of recruiting for active measures—including ‘wet work’ in the United States, and the threat AI may pose to authoritarian regimes.