A video that appears to show the president is circulating on social media after a very public line erupted between President Donald Trump and Elon Musk, who announced a ban on Tesla production in the United States.
But President Trump didn't say this, and the clip was edited using artificial intelligence (AI).
In the video, President Trump said: “Today, I'm here to announce that we'll be banning all Teslas production in the United States right away. Everyone knows that a few days ago Elon stabbed me in the back and became obsessed with his platform.”
Tesla is an electric car company led by Musk. After the disagreements were made public, there were reports that investors sold stocks in the company, but no credible reports of the president have found that he would ban production in the United States.
Where did the video come from?
The earliest example of the clip was posted on Instagram on June 7th by an account called @DangerousAirturns.
Some versions shared on social media include the @dersuralai2 tiktok handle in the watermark that shares a similar name to your Instagram account. The Tiktok video has since been deleted, but it still appears in Google search results. The Facebook version under the Thedangerousai handle is still live.
When contacted, the Dangersai YouTube channel earned credits for using AI to create videos.
It appears to be an edited version of a real clip of Trump. Using Google lenses, Full Fact followed the original clip into a video of him speaking in the oval office next to Musk on May 30th. The modified clip appears to be inverted horizontally, so in the original video, the US flag pin badge appears on President Trump's right rappel rather than on the left.
President Trump has not mentioned the original video banning Tesla production in the United States.
It also suggests that the audio that President Trump is speaking on the viral video is Deepfark.
Ritasin, a scientist and expert in human voice profiling technology at the Center for Voice Intelligence and Security at Carnegie Mellon University in the US, said the complete fact that audio exhibits “classic features where AI is generated,” such as “no impocaric breath sounds” and “inconsistent” speech patterns.
She explained: “Audio shows that unique vocal patterns, which are usually consistent on a single speaker across different sounds, are not uniform in this recording. What's more, the way the sounds mix and influence each other's frequency responses differ in an unnatural way between different sounds and different instances of the same sound.”
The versions circulating on social media are also blurry around the mouth. As Dr. Dominic Rees, an associate professor at Generative AI's Reading University, said before, this is a sign that Deepfark is not lip-synced in the footage.
As technology advances, deepfakes become more and more realistic and difficult to distinguish from real footage. A guide to discovering Deepfake videos and AI audio provides practical tools to determine if they are authentic or not.
