Can you spot the deepfakes? Our democracy may depend on it
It would be a shame if the upcoming presidential election were won on a lie.
However, as AI technology evolves, the creation of highly believable fake content has the potential to spread rapidly and mislead a deeply divided voter base during a close election.
So far, there have already been meme deepfakes posted by Elon Musk, as well as fake AI ads of celebrities endorsing various political figures.
As voters and users of social media, we have a responsibility to be smart consumers of content, regardless of how posts are being moderated by social media platforms.
Distinguishing fact from AI fiction on social media is becoming increasingly difficult. Despite efforts from major technology companies — including the signing of a pledge to moderate AI content that can potentially interfere with the election — deepfake voices and images still run rampant, leaving users with an AI Pandora’s box.
“The biggest challenge isn’t just removing harmful content — it’s identifying what qualifies as harmful in the first place,” Michael Fleischman, a former TikTok content moderator, said. “With the rise of AI-generated content, misinformation can look incredibly convincing, making it harder for both platforms and users to tell what’s real and what’s not.”
Elections are particularly vulnerable. In 2016, Russia ran a potent disinformation campaign to support Donald Trump by creating a troll farm, where millions of social media bot accounts were created to post pro-Trump content. Now, with AI able to deepfake voices and images, uncanny videos created by domestic and foreign bad actors can spread fake information rapidly.
The rise of AI presents both a challenge and a potential solution to misinformation. Social media platforms such as TikTok are using AI to detect and combat fake AI content. But fake content still leaks through. Fleischman describes it as being a game of “whack-a-mole,” with new AI content being moderated by improved AI detectors, which is then circumvented by newer artificially generated content.
“AI is learning to combat AI-generated misinformation, but it’s an ongoing battle,” Fleischman said. “We were teaching our system to recognize patterns, but as it gets better, the tools are evolving, but they need constant refinement to keep up with new types of content we see every day.”
This cat-and-mouse chase is ongoing as the latest detectors such as deepware.ai can detect fake AI videos of Russian president Vladimir Putin from two years ago, but are unable to detect fake AI videos of U.S. official Matthew Miller pushed by Russia five months ago, nor a recent meme deepfake posted by Elon Musk of Kamala Harris’s campaign. Newer deepfakes are harder to detect because they use the latest AI advancements that AI detectors have not been trained on, and can draw from a vast pool of celebrity video and audio data on the internet. As AI technology improves, it has become even easier to create precise deepfakes.
Therefore it is up to us, the users, to judge the content that we see. Users must play an active role in discerning what is real and what is not. With pundits predicting this upcoming election between Trump and Harris to be the closest in a generation, it is more important than ever that we help police the content that we see, and ensure our election is fair. There are four main principles that Fleishman advises users to follow when judging content:
- Look for extreme claims when supposed facts are presented.
When content presents itself as factual but sounds too extreme, treat it with skepticism. Watch out for absolute terms like “always” or “never.” These are often signals of bias or an attempt to manipulate the audience’s perspective. - Be very wary of news media or content presenting facts that play on emotions.
If content tries to provoke a strong emotional response — whether it is fear, anger, or outrage — consider why. Genuine information usually does not need to manipulate emotions to be convincing. - Always take a step back when looking at a post and see what the general message is.
Ask what the overall message is and whether it seems credible. Think of the broader reason for a particular post, and what the intentions may be. - Make sure to always report suspicious content and spread the word to your friends. Better be safe than sorry.
If you think a piece of content is fake based on the principles above, then report it so it can be flagged for other users. It is our responsibility to help stop the spread of misinformation by flagging what is fake to the public.
The responsibility to discern truth from falsehood is a shared one in a democratic society. While social media platforms are on the front lines, users must also be on guard while navigating political content. The fate of our democracy depends on the choices we make today.
