The Rise of AI Deepfakes in Elections

Artificial intelligence is revolutionizing the landscape of election disinformation, with the ease of creating convincing fake content becoming more accessible to anyone with a smartphone and a mischievous mind. Gone are the days when producing fake photos, videos, or audio clips required teams of experts with technical skills and a hefty budget. Thanks to companies like Google and OpenAI offering free or low-cost generative AI services, the creation of high-quality “deepfakes” now only requires a simple text prompt.

The impact of AI deepfakes on elections has been felt across Europe and Asia, serving as a cautionary tale for the more than 50 countries gearing up for elections this year. Henry Ajder, a leading expert in generative AI based in Cambridge, England, noted that confusion over the authenticity of content is becoming more prevalent. The concern now lies not in whether AI deepfakes could influence elections, but in how significant their impact will be.

As the U.S. presidential race intensifies, FBI Director Christopher Wray has issued a stark warning about the escalating threat posed by generative AI, stating that foreign adversaries could easily engage in malicious influence operations. With AI deepfakes, a candidate’s image can be tarnished or polished, steering voters towards or away from certain candidates, or even discouraging them from voting altogether. However, the most significant danger experts foresee is the erosion of public trust in what they see and hear.

Recent examples of AI deepfakes include a video of Moldova’s pro-Western president endorsing a pro-Russian party, audio clips of Slovakia’s liberal party leader discussing vote rigging, and a video of an opposition lawmaker in Bangladesh, a conservative Muslim-majority nation, donning a bikini. The sophistication of AI technology makes it challenging to trace the origins of these deepfakes, raising concerns about the lack of mechanisms to counter the deluge of disinformation.

In Moldova, President Maia Sandu has been a frequent target of AI deepfakes, with one circulating video depicting her endorsing a Russian-friendly party, a ploy believed to be orchestrated by the Russian government. Similarly, China has been accused of leveraging generative AI for political manipulation, as seen in Taiwan, where a deepfake video stirred concerns about U.S. interference in local politics.

The audio-only deepfakes pose a unique challenge in verification, as they lack the visual cues that signal manipulated content. In Slovakia, audio clips resembling the voice of the liberal party chief circulated widely before parliamentary elections, discussing raising beer prices and vote rigging. The deceptive nature of these deepfakes underscores the challenge of distinguishing truth from fiction, especially in regions where media literacy is low.

As the world grapples with the proliferation of AI deepfakes, authorities are racing to implement safeguards. The European Union has mandated special labeling of AI deepfakes starting next year, while tech giants like Facebook are taking steps to label deepfakes on their platforms. However, the decentralized nature of platforms like Telegram poses a challenge in curtailing the spread of deepfakes.

The use of generative AI in political campaigns is on the rise, with campaigns leveraging AI to enhance a candidate’s image. In Indonesia, the team behind presidential candidate Prabowo Subianto’s campaign used a mobile app to deepen connections with supporters. The evolving landscape of AI deepfakes calls for global cooperation to combat misinformation and safeguard the integrity of elections.

In a world where disinformation can be easily manufactured, the threat to democracy looms large. Efforts to counter AI deepfakes must strike a balance between preventing malicious manipulation and preserving freedom of expression. As we navigate this complex terrain, the need for vigilance, transparency, and collaboration has never been more critical to safeguard the democratic process.

AP News

About The Author

Scroll to Top