Experts fear AI deepfakes can deceive voters in Pakistan, India, and other Asian nations
With elections scheduled in Pakistan, India, and other countries in Asia, the spread of misinformation, especially through deepfakes, is a growing concern. Deepfakes are videos or audio recordings created using artificial intelligence (AI) that are designed to appear authentic but are actually fabricated. This technology poses a significant threat to the democratic process, as it has the potential to manipulate and influence voters.
Divyendra Singh Jadoun, the founder of The Indian Deepfaker, a company specializing in AI-based visual effects and voice clones for film and television, has been approached by politicians requesting deepfake videos for their election campaigns. While the demand for such content is high, Jadoun is hesitant to engage in this practice.
The technology to create deepfakes is so good now, it can be done almost instantaneously, with very little effort – and people cannot tell if it’s real or fake, said Jadoun. There are no guidelines on deepfakes, and that’s worrying, as it has the potential to influence how a person votes.
Instances of deepfakes being shared on social media platforms are on the rise, particularly in countries like India, Indonesia, Bangladesh, and Pakistan where elections are imminent. Instagram reels of Indian Prime Minister Narendra Modi singing in regional languages and TikTok videos of Indonesian presidential candidates speaking fluent Arabic have gone viral, despite being AI-generated and posted without disclosure.
In response to this alarming trend, Indian Prime Minister Narendra Modi has emphasized that deepfake videos are a big concern, and authorities have warned social media platforms about losing their safe-harbor status if they fail to take action against such content. Similarly, in Indonesia, where over 200 million voters will cast their ballots, deepfakes of all the presidential candidates are circulating online, potentially impacting the election outcomes.
Nuurrianti Jalli, an assistant professor at Oklahoma State University’s media school, cautions that AI tools like deepfakes can significantly influence voter perceptions and behavior by targeting individuals with disinformation and spreading false narratives on a large scale and at an unprecedented speed. In environments already plagued by misinformation, AI-generated content further skews public perception and influences voting behavior.
Last year, deepfake images and videos created using generative AI tools made headlines in countries like New Zealand, Turkey, Argentina, and the United States, raising concerns about their impact on elections. The use of AI in disseminating disinformation makes it faster, cheaper, and more effective, according to a report by US non-profit organization Freedom House.
In Bangladesh, where Prime Minister Sheikh Hasina is set to secure her fourth consecutive term, deepfake videos of female opposition politicians have emerged. Despite being quickly debunked, the circulation of even poor-quality deepfake content continues to mislead people. Sayeed Al-Zaman, an assistant professor of journalism at Bangladesh’s Jahangirnagar University, warns that in a country with low levels of digital literacy, deepfakes can be powerful carriers of political propaganda.
Meanwhile, in Pakistan, where an election is scheduled for February 8, incarcerated former Prime Minister Imran Khan used an AI-generated image and voice clone to address an online election rally in December. The event garnered over 1.4 million views on YouTube and was attended virtually by tens of thousands. While Pakistan has drafted an AI law, digital rights activists are concerned about the lack of safeguards against disinformation and the protection of vulnerable communities, including women.
There is a growing recognition that platforms like Facebook, Instagram, WhatsApp, and YouTube are struggling to keep up with the spread of deepfakes. Social media giant Meta, which owns these platforms, aims to remove synthetic media that could mislead viewers when the manipulation is not apparent. Google, which owns YouTube, requires creators to disclose altered or synthetic content but acknowledges the need to inform viewers about such content through labels.
However, countries like India, Indonesia, and Bangladesh have recently enacted stricter laws to regulate online content and hold social media sites accountable for misinformation. Consequently, platforms are exercising caution and self-censorship due to fears of penalties. Raman Jit Singh Chima, Asia policy director at advocacy group Access Now, warns that this lack of responsiveness and proactivity from platforms in the face of deepfake proliferation is a dangerous sign.
As the general election approaches in India, Divyendra Singh Jadoun, having previously declined to create deepfake campaign videos for state elections, is now preparing to make personalized video messages from politicians for party workers. These messages will be sent via WhatsApp and have the potential to sway voters as party workers forward them to their friends and family. To prevent misunderstandings, Jadoun plans to add a watermark indicating that the videos are AI-generated.
The prevalence of deepfakes continues to grow, with an estimated 500,000 video and voice deepfakes shared on social media globally in 2023, according to DeepMedia, a company specializing in detecting synthetic media. While platforms attempt to tackle this issue, the danger deepfakes pose to the integrity of elections and the democratic process cannot be overstated.
Disinformation, amplified through AI-generated content, has the power to influence voting behavior, sway public opinion, and impact the outcomes of elections. As the world pays close attention to the US election, it is crucial to apply similar standards and efforts globally to combat the threat posed by deepfakes.
