There has been a surge in deepfakes created by artificial intelligence on social media, causing widespread concern about manipulated media ahead of the upcoming US election cycle. Fake videos and images, including pornographic images of singer Taylor Swift, robocalls of US President Joe Biden’s voice, and videos of dead children and teenagers have gone viral, despite being completely falsified.
While misleading audio and visuals created using artificial intelligence are nothing new, recent advances in AI technology have made them easier to create and more difficult to detect. The highly publicized incidents have raised concerns among lawmakers and citizens alike.
White House press secretary Karine Jean-Pierre expressed alarm at the circulation of false images on Jan 26 and stated that the administration would take necessary action to address the issue.
The spread of AI-generated fake content on social networks has also highlighted the limitations of platforms’ ability to police them. Explicit AI-generated deepfaked images of Swift gained millions of views on X, a platform owned by Elon Musk, before being removed, indicating the potential for such images to go viral before any action is taken to stop them.
Cracking down on fake content
AI expert Henry Ajder emphasized the need for companies and regulators to take measures to stop the spread of manipulated content. He noted the importance of creating friction in the process from the idea to the creation and sharing of such content.
Following the episode involving Swift, there was public outrage on X with the phrase “protect Taylor Swift” trending on the platform. This is not the first time the singer’s image has been used in explicit AI manipulation, but it has sparked a new level of public outcry.
The top 10 deepfake websites referenced “Taylor Swift” in about 1,000 videos by the end of 2023, and many of these videos are accessible through a simple Google search. However, the process of requesting removal of deepfake content from Google has been likened to a game of whack-a-mole.
It is important to note that almost 500 videos referencing Swift were hosted on Mrdeepfakes.com, and the site received 12.3 million visits in December 2023
Targeting women
According to AI expert Henry Ajder, the ease of creating explicit content using AI is disturbing and disproportionately affects women and girls, regardless of their social status.
As of Friday afternoon, explicit AI-generated images of Swift were still present on X, despite the platform’s policy against non-consensual nudity, sparking concerns about the regulation and removal of such content.
Surging numbers
Research from independent analyst Genevieve Oh has shown a significant surge in pornographic deepfake videos since 2020, with the top 10 sites offering this content hosting over 114,000 videos, many of which targeted Swift.
Digital forensics expert Heather Mahalik Barnhart highlighted the impact of deepfake content on individuals, even when it is fake, noting the potential psychological distress it can cause to the victims and their families.
Concerns about the use of deepfake audio or video in upcoming elections have also been raised, as evidenced by a deepfake audio message of Biden spread ahead of the New Hampshire presidential primary election.
Difficult to detect
The spread of deepfake videos and images on social media poses a significant challenge due to the lack of reliable detection capabilities. Even prominent public figures like Swift and Biden find it difficult to address and remove such content before it has already caused damage.
There is also a lack of concrete laws in the US addressing deepfakes, particularly those of a pornographic nature, making it difficult for victims to hold creators accountable.
Few laws
Despite the lack of federal laws prohibiting deepfakes, the administration is working with AI companies to develop unilateral efforts to watermark generated images and address online harassment and abuse. Congress has also begun discussing legislative steps to protect celebrities’ and artists’ voices from AI usage, but there are currently no protections for private citizens.
There has been public outcry over the use of AI technology to create explicit content, which has sparked widespread concern about the potential impact on individuals and upcoming events such as elections. Improved detection technologies and legislative efforts are necessary to address this growing issue.
– Bloomberg
