Disclaimer: Reality Defender seeks to combat mis/disinformation and the harms caused by dangerous deepfakes and misused AI. We do not, as an organization, take a political stance, and view deepfakes and related ills as an issue that impacts all people, regardless of political or socioeconomic affiliation (among other qualifiers).
On Sunday, former president and current Presidential Candidate Donald Trump took to his social media platform, Truth Social, to share AI-generated images of women wearing “Swifties for Trump” t-shirts, along with an AI-generated diffusion image of Taylor Swift dressed as Uncle Sam declaring “Taylor Wants You.” By combining these posts with a reply of “I accept,” the former president appeared to imply a valuable endorsement from Swift that he has yet to receive.
Another post also seemed to suggest that Swift’s fandom has become supportive of his campaign after the foiled terrorist plot targeting the singer’s concerts in Austria. The former president’s use of deepfake images comes after he has warned against the nefarious use of AI and accused the Harris campaign of utilizing deepfake technology to create a false impression of the size of her rally crowds.
The Importance of Labeling AI-Generated Content
As of this writing, the former president did not acknowledge that the images were AI-generated, and the Truth Social platform did not add any warning labels to the post. The incident has been widely reported as disinformation by news outlets around the country, with reporters suggesting that the former president, who has long sought Swift’s endorsement and the votes it could bring, attempted to utilize AI to create a false narrative that the endorsement is forthcoming or already in place.
In instances when public figures with mass followings utilize AI in ways that may create misleading narratives — intentionally or not — it is especially important for all deepfakes to be properly labeled by social media platforms on which the images are distributed. In this instance, Truth Social could have avoided any misunderstanding by labeling the media posted by the former president from the outset.
What Reality Defender Found
The Reality Defender platform detected that some of the images were likely generated using AI.
Regulating AI in Politics
Given how easy it is to create misleading narratives with deepfakes, legislators must hold politicians to a higher standard by establishing clear rules for how AI-generated media can be utilized by public servants, whether the use is for the purpose of campaigning, announcing endorsements, or communicating with the public while in office.
The ELVIS Act adopted by the Tennessee legislature, which makes it illegal to utilize a person’s AI-generated likeness for advertising or political endorsement without consent, could become a model for the rest of the country — and could even come into play in this specific instance.
More than ever, it is important that social media platforms that host political accounts — and thus serve as a part of the news media ecosystem — adopt strict policies that detect deepfakes like these and subsequently label all and any AI-generated content to bring needed clarity to their users and avoid the spread of disinformation. Digital platforms needn’t rely on politicians to make the initial disclosure about synthetic media. By adopting deepfake detection tools as part of their content dissemination workflows, platforms can detect AI-generated media on their own, and thus remain fair and objective while navigating the complex landscape of political free speech.
