I listened to an interesting podcast on the implications of generative AI in elections last week that I thought people in this class might enjoy. While the podcast makes note of the benefits of generative AI to elections, specifically around generative AI’s capacity to reduce campaign costs, it mostly focuses on the threats that generative AI poses to electoral processes across the globe. Among the potential threats is the creation of audio and/or video deepfakes of candidates, which can not only be easily made by anyone using generative AI, but could have serious influence on election results. For example, the podcast hosts discuss how an audio deepfake of a Slovakian candidate where he alleges that he rigged the election and that he would raise the price of beer if elected could have swayed the election results. Notably, this deepfake was released days before the election day and, although unconfirmed, possibly contributed to his defeat given that the two leading candidates had been polling equally in the weeks preceding the election.
Given that it is an election year for the US (and Canada’s next election is in 2025), the question hangs as to how countries should address the issue of AI-generated content of candidates that might filter onto social media, or whether the onus should fall on social media companies rather than legislatures. As the podcast notes, five US states have enacted regulations around AI-generated deepfakes with varying penalties and many more have legislation that is being debated. Although this form of legislation might function to deter individuals from creating and disseminating deepfakes, these laws cannot address the potential damage wrought through deepfakes that do circulate, particularly if they are released just days before an election and with little time to dispel the legitimacy of the audio.
As we discussed in class with the Taylor Swift deepfakes, policing of deepfakes might fall on individuals who can try to bring this content to the attention of social media platforms more quickly than these platforms seem to be responding. While individuals can assist in getting content off platforms, candidates who are affected will likely face challenges in redressing the harms they experience, and copyright law likely offers little protection for issues of this nature.
