The Wired AI Elections Project tracks how generative AI is impacting elections globally, with a focus on deepfakes, AI-generated content, and AI candidates. It includes a geospatial map showing instances of AI use in elections, categorized by region and type of use.
Deepfakes are being used in both sanctioned and unsanctioned ways, such as the Biden robocall discouraging voters and Imran Khan's authorized deepfake declaring victory while in prison. They are often emotionally resonant, even if people know they are fake, and are harder to detect in audio form, especially in regions like South Asia.
The liar's dividend refers to the idea that when everything can be falsified, nothing is real. This concept is amplified by AI, as people may question the authenticity of even real evidence, leading to a blurring of shared reality and truth in political discourse.
AI candidates, like A.I. Steve in the UK and Yas Gaspadar in Belarus, are being used to represent political dissidents who cannot run in person due to safety concerns. These AI avatars allow for virtual representation and engagement with constituents, bypassing physical limitations.
Platforms struggle with detecting AI-generated content due to lack of transparency in detection systems and inconsistent labeling practices. Watermarking and hashing technologies could help, but they require universal adoption and cooperation among platforms, which is currently lacking.
In countries like India and Pakistan, AI is being used for deepfake audio messages and personalized outreach via WhatsApp. In Indonesia, AI avatars are used to create viral campaign content on TikTok, reaching billions of views and influencing voter perception.
AI can influence voter turnout and perception by creating personalized outreach that makes voters feel seen, even if it's automated. In some cases, AI-generated content can go viral, significantly impacting a candidate's popularity, especially among younger voters who may not have historical context for certain political figures.
The broader implications include a potential erosion of trust in institutions and shared reality, as well as the need for increased investment in trust and safety measures on platforms. The lack of these measures could lead to more AI-generated disinformation and further polarization.
Each Sunday, TED shares an episode of another podcast we think you'll love, handpicked for you… by us. Today, an episode of The TED AI Show. 2024 was the biggest election year in modern history, with over 50 countries going out to the polls across the globe. And artificial intelligence has fully seeped into global politics – from deepfakes to AI bots that can ingest thousands and thousands of documents to make policy decisions. The TED AI Show host Bilawal Sidhu talks with journalist Vittoria Elliot, who’s been leading on WIRED’s AI Elections Projects, to discuss how AI is reshaping the political landscape in surprising ways. The two explore the good, the bad, and the downright bizarre – and share what the U.S. can learn from other countries to adapt and critically engage with "the new normal." Hosted on Acast. See acast.com/privacy) for more information.