India's Billion-Voter Election Met the Deepfake
During India's 2024 general election, artificial intelligence revived a politician who had been dead for six years and put words in the mouths of Bollywood stars. Surveys, the Election Commission and a Meta-funded WhatsApp tipline all scrambled to keep pace.
India's 2024 general election was the largest exercise of its kind ever held, with close to a billion registered voters and polling staged across seven phases from April to June. It was also, researchers and fact-checkers say, a stress test for what generative artificial intelligence can do to an information environment at scale.
One of the most striking uses came from the south. The Dravida Munnetra Kazhagam, the party that governs Tamil Nadu, used AI to resurrect its late leader, M. Karunanidhi, who died in 2018, “using lifelike videos of the former movie writer and veteran politician at campaign events,” Al Jazeera reported. Karunanidhi, a former chief minister who had led the party for decades, appeared in the clips wearing his trademark dark glasses and yellow shawl, addressing the living from beyond the grave.
Fabricated endorsements
The deception reached beyond politicians. In April 2024, around the start of the first phase of voting, deepfake videos of the Bollywood actors Ranveer Singh and Aamir Khan spread on social media, appearing to show them criticizing Prime Minister Narendra Modi. A spokesperson for Ranveer Singh said the actor's team had gone to the police: “Yes, we have filed the police complaint and FIR has been lodged against the handle that was promoting the AI-generated deepfake video of Ranveer Singh,” the spokesperson told Outlook Business. Both actors' images were used without consent.
Surveys suggested exposure was widespread. A study by the cybersecurity firm McAfee, based on 7,000 respondents surveyed across seven countries in January and February 2024, found that about 75 percent of Indians had encountered deepfake content, and that nearly one in four — 22 percent — said they had come across political content they later discovered was a deepfake, as reported by The Tribune.
Regulators and fact-checkers respond
As complaints mounted, the Election Commission of India issued an advisory on 6 May 2024 directing political parties to take down deepfake audio or video within three hours of becoming aware of it, and to identify and warn whoever created the material, TechCrunch reported. In its letter to the parties, the commission warned that “the use of such manipulated, distorted, edited content on social media platforms has the potential to wrongfully sway voter opinions, deepen societal divisions, and erode trust in electioneering process,” as quoted by the fact-checking outlet BOOM.
Ahead of the vote, a coalition of fact-checkers had opened a public channel for suspect material. The Deepfakes Analysis Unit, set up under the Misinformation Combat Alliance and funded by Meta, launched a WhatsApp tipline in March 2024, less than a month before the first phase of polling, the World Economic Forum reported. It invited users to submit suspicious audio and video for analysis in English, Hindi, Tamil and Telugu, and reviewed hundreds of files during the campaign.
A template for future votes
Researchers who tracked the campaign say India's scale made it an early preview of AI-shaped elections elsewhere. Cheap synthetic voices and faces let campaigns and unnamed operators flood messaging apps faster than any single fact-checking unit could respond, and much of the most viral material was crude edits rather than sophisticated forgeries. The durable question, analysts argue, is whether disclosure rules and rapid-takedown windows can be enforced across platforms that reach hundreds of millions of voters in a matter of hours.