The use of deepfakes, or manipulated videos and audio recordings created using artificial intelligence, is on the rise as the 2024 US presidential election approaches. Such videos, also known as ‘diphs’, are a powerful tool for manipulating public opinion and discrediting rivals. Examples of diphs include one in which Hillary Clinton appears to support Republican Rona Detis, and another in which President Joe Biden insults transgender people.
Creating diphs has become easier and cheaper thanks to new “generative AI” tools such as Midjourney, which allow the synthesis of realistic images and voices. By 2023, the number of deepfakes on the internet is predicted to triple in comparison with last year, with some estimates suggesting that around 500,000 video and audio deepfakes will be published annually.
One report suggests that voters will find it difficult to distinguish between real and fake content. “And one can imagine how supporters of Trump or Biden can use this technology to force the enemy to appear in the worst light,” warns Darrell West, senior researcher at the Brookings Institution. Despite efforts by social media platforms such as Facebook, Twitter and YouTube to remove diphs, their effectiveness remains low.
Some politicians are using generative AI to gain an advantage in election campaigns, including the Republican National Committee, which created a deepfake video showing a dystopic vision of America under Biden. With further investment, these technologies could easily sway the results of elections, leading to concerns that they could be used to manipulate public opinion and push an agenda.