Beyond Deepfakes: The Generative AI Threat to 2024's Elections
How platforms like TikTok become battlegrounds for AI-powered propaganda
I recently co-authored a by-invitation article in The Economist.
The gist of the article is how generative AI is changing the playbook for disinformation just as we hit a historic year of elections. The number of countries having elections in 2024 ranges from 64-80 depending on whether one wants to include the exercises in rubber-stamping. This is happening just as advanced generative AI has become widely available.
In the article we focus specifically on examples from Taiwan, which had its Presidential election in January, and who has long been targeted by influence operations from China.
One of the points I hope comes out in the article is that disinformation influence operations are following the shift of online attention to short-form video platforms like TikTok.
Generative AI tools for video-editing make it cheap to produce compelling video for these platforms that tell any narrative you want. Here is a video I made to promote our article using a combination of “deepfake” tools, and tools for using AI to generate B-roll. It only took me about 5 minutes.
I think most journalists aren’t really grasping the implications of this problem. They understand “deepfakes” — that you can use AI to generate fake photo and video records of events that did not occur — the kinds of things fact-checkers could debunk.
But you can use generative AI tools to create more than fake records of events. You can use them to turn any narrative into content that is more engaging, sharable, and viral. These include false or deceptive narratives that lean more into half-truths, stereotypes, predictions, and identity politics, rather than outright falsehoods. Imagine how the advances in text-to-video tools (e.g., OpenAI’s Sora) could be used to promulgate conspiracy theories that is particularly salient to some segment of an electorate.
This article was part of some ongoing research I’ve taken on with a colleague Madeleine Daepp. Links to the article are below.
But you don’t need us to tell you that 2024 is going to be a wild year.
Go Deeper
Daepp, M.I.G. and Ness, R.O., 2024. By Invitation | False narratives: Video will kill the truth if monitoring doesn’t improve, argue two researchers. The Economist.
Archived version (no paywall)
Meeting the moment: combating AI deepfakes in elections through today’s new tech accord. Microsoft Blog
Seeing is not believing—deepfakes and cheap fakes spread during the 2024 presidential election in Taiwan | 台灣事實查核中心 (tfc-taiwan.org.tw)