Scammers have always tried to trick us into handing over our hard-earned money and personal information. But generative AI is making it even easier for them. Now, with the click of a button, they can create pretty decent content that looks and sounds just like us. This is called AI impersonation, and it’s a growing problem.
Celebrities, public figures and politicians are all high-profile deepfake targets. As we approach what is likely to be consequential and controversial midterm elections in November, we need to prepare now.
AI-generated or synthetic media is the new norm, thanks to tech companies like OpenAI, Google and Meta. AI image and video generators can create and edit content easily, no advanced Photoshop skills required. AI makes it easier to make more fake content and quickly distribute it widely. Bobby Ford, chief strategy and experience officer at Doppel, told me that this problem can be thought of as the three V’s: volume, velocity and vectors.
“Before, it would take a threat actor a certain amount of time, as well as resources, in order to create that deepfake. Take that same amount of time, and it’s done at much faster clips, so the volume (and velocity) increases,” Ford said. New vectors, like video, audio and email, are also available as AI tools and have become increasingly multimodal, he said, “so we’ve seen those three V’s increase.”
But the goal of synthetic media isn’t always to make a flawless imitation of a politician or public figure. It’s often used to make us believe the fake content is real — to manipulate our emotions and get us to override our better judgment. That ability to influence our beliefs matters a lot, especially in politics.
AI is already embedded in politics
During the 2024 primary presidential election, thousands of voters in New Hampshire got surprise robocalls from Joe Biden, urging them not to vote the night before the election. Except, it wasn’t Joe Biden. It was an AI deepfake, and voters struggled to tell the difference. The Biden campaign tried to let people know it was fake, but the damage was already done.
Political campaigns “may not have the tools for cybersecurity, but they’re also out there in the world using all the platforms that the bad actors use in order to reach the same group of people,” said Michael Kaiser, president and CEO of the nonprofit Defending Digital Campaigns. “That makes it extremely difficult to operate in this environment.”
The goal is to erode trust and sow chaos during election season and beyond. In July, Sen. Mitch McConnell’s office released a picture of him during his recent hospitalization. Users on social media immediately began questioning whether it was real. Many folks shared the theory that the photo was either old or edited with AI tools. Digital forensics experts found no evidence that it was old or digitally altered. But we live in a world where that’s one of our first thoughts.
“If that photo had been released two years ago, it would have taken me probably a good seven or eight hours to doctor that photo,” Ford told me. “But now, with generative AI, all I need to do is copy that photo, drop it into an app, give it some plain language commands and within 30 seconds, I can have 10 different versions of that same photo, and then I can send those 10 different versions to my friends.”
It’s that speed and ease that help bad actors flood the internet and our phones with this spammy or scammy content. But while the tech may be new, the tactics they use aren’t. They can intentionally craft the messages to have a strong urgency so you don’t pause to think about a reply. For example, if you get a text telling you that your polling location has been changed right before election day, you might not stop to question whether it’s true before you click a malicious link.
“The point of this media is the manipulation of your emotions,” Kaiser said.
So one of the best ways to fight back is to take a breath, step back and think before you do anything rash — and ask why this message is being sent to us. But that’s easier said than done sometimes.
“Now I don’t need to do a bunch of research on (people) in order to target (them),” Ford said. “I could go to some app and have it write either an email or have it develop some synthetic voice based on open-source intelligence it can gather on you — that’s the difference.”
