AI: The new campaign tool? AI reshapes political campaigns in Tamil Nadu; Emergence of deepfakes raises ethical concerns

Six months ago, an AI-generated video circulated on social media showing Tamil Nadu’s first chief minister C N Annadurai ‘urging’ actor-politician Vijay to lead with his iconic line “thambi vaa, thalaimai yerka vaa” (come brother, take the lead). The clip of the late politician was among the first widely reported uses of generative AI in Tamil Nadu ahead of the 2026 Assembly election. Within hours, it drew criticism and questions of ethics from rival parties; this immediate political response that set the pattern for how AI would enter state politics.color:black'>DMK, which traces its ideological lineage to Annadurai, responded within a day through its IT wing by releasing an AI-assisted video invoking Periyar and Annadurai, which was amplified by party functionaries accusing rivals of appropriating Dravidian legacy using AI. “This made it clear that the problem was with what the video showed, not with the use of AI itself. That exchange showed how AI had moved from being a fringe experiment to a normal campaign tool,” says political observer R Chandrasekaran.color:black'>That episode became a turning point. “Parties realised AI could turn a single speech or image into multiple social media posts far faster than the WhatsApp messages, posters and manually edited videos they had relied on earlier,” he says.color:black'>In the following weeks, AI-assisted video production became routine across parties. Short clips with automated subtitles, AI-generated backgrounds and music began appearing on the official handles of DMK, AIADMK and BJP’s Tamil Nadu unit.While BJP’s Tamil Nadu unit mostly adapted centrally produced AI content for Tamil audiences by replacing national leaders with local faces, AIADMK relied on animated videos such as its viral ‘Uruttu Kadai’ halwa clip accusing the DMK of unfulfilled promises. DMK countered with videos titled ‘Pathu Tholvi Palaniswami’, targeting EPS over repeated defeats. “Once a party releases a provocative AI clip, we have to quickly respond with similar content to avoid falling behind online,” says Kovai Sathyan, AIADMK’s IT wing state president.Things got murkier when AI use went beyond campaign messaging. Deepfake videos of leaders prancing around in comic settings began to proliferate. “This soon gave way to more serious misuse, with obscene and defamatory deepfake images and videos targeting leaders from various parties. Clips pasted faces onto explicit imagery or created fake audio meant to embarrass leaders,” says R Gopinath, a political science academic.Parties then began to file police complaints. Police have made arrests in some cases, but tracing creators is often difficult as the content is made using easily available tools such as Stable Diffusion for images, ElevenLabs for audio, and Reface or DeepFaceLab for deepfake videos, often from outside the country, say party insiders.But with parties now using AI for more than just sledging, analysts say this could be the first full campaign cycle in the state where it plays a central role rather than remaining a novelty. DMK has announced an AI-enabled portal to collect and sort public suggestions for its 2026 manifesto. “AI is also being used to read data points of different populations and translate them into marketing and communication. We can replicate speeches, send personalised calls, and automate WhatsApp videos for specific occasions such as birthdays and anniversaries,” says Jai Pratap, director of political consultancy firm Political Edge, who has worked on multiple campaigns for prominent parties in Tamil Nadu.Jai says AI is increasingly becoming the “think tank” behind campaigns. “Earlier, you needed teams brainstorming ideas for days. Now, you prompt and get multiple campaign ideas instantly,” he says. For example, Jai says, when one of the political parties wanted to mobilise the youth in Chennai, AI asked them the economic, social, and cultural perspectives of the location they aimed to target. “It designed a campaign which worked towards addressing the issues people faced in this particular area.” In another instance, Jai says, when AI was tasked with designing a campaign for an MLA candidate, it produced a strategy centred on exploiting the sitting MLA’s administrative shortcomings. “It drew up a detailed list of strengths and weaknesses, based on multiple reports, that the current MLA was perceived as unapproachable and inaccessible to the public. AI then suggested building the campaign around positioning our candidate as someone the public could rely on.”Political Edge is also testing AI filters on Instagram and Facebook that let party workers and supporters easily share photos and videos with party symbols and messages. “If I create one filter with the party flag and message, and have consent from one lakh cadre accounts, a central system can post the content at scale,” he says.Divyendra Singh Jadaoun, founder of The Indian Deepfakes, says the evolution has been rapid. His company has developed AI software for many political campaigns, including AI voice-cloned calling systems that allow two-way communication with voters. “Voters get a call that sounds like the political candidate who is contesting in the elections, and it listens to their concerns and responds in real time,” he says. The data is also saved into a system that allows analysts to present their views and issues to the candidate to refine their strategies continuously. Jai says that recently, a political consultant from Chennai approached his firm for a similar system that will soon kick off.In Andhra Pradesh, Jadaoun’s firm created an AI avatar of Y S Rajasekhara Reddy that could “talk” to people in real time, while a similar conversational avatar was built for the President of Suriname in 2025. But key campaign decisions still depend on traditional surveys, field workers and booth-level feedback, with AI tools discussed but not yet trusted for major decisions,” say strategists, primarily because local factors such as caste equations, neighbourhood issues and personal influence are hard for off-the-shelf AI tools to capture. Also, parties worry that voter data, especially after the Special Intensive Revision (SIR) exercise, remains incomplete or outdated. “For now, parties see this technology as something that boosts their work, not replaces organisation on the ground. This gap may narrow before 2026,” says a DMK war room staffer.“Experiences elsewhere show where the use of AI could head,” says D Raj, who has worked with political consultancy firms. “AI-generated robocalls mimicking Joe Biden’s voice were used in the 2024 US elections. Similarly, in Tamil Nadu, parties could move from one-way IVRS (recorded voice) calls to interactive voice systems in leaders’ voices that answer voter questions.”AI tools can be used to understand voter behaviour, as the Democratic National Committee (DNC) did in the US with Matchbook to identify persuadable voters and OpenField to track campus canvassing, adds Raj.(with inputs by Yazhiniyan)“Detection of fakes has become difficult,” says Divyendra Singh Jadaoun, founder of The Indian Deepfakes. “What once took days and significant resources can now be created in minutes for under five dollars, largely due to open-source tools. These videos are hyper-realistic. Sometimes I get confused whether it’s real or not.”Jai Pratap, director of political consultancy firm Political Edge, says regulation and detection systems are lagging behind innovation. “We always use a watermark so people know it’s AI-generated. We don’t give these tools and software to the parties, because it can be misused. If you prompt it to blackmail or clone an opposition leader’s voice, it may do that,” says Jadoun.Jai says a few parties also make requests involving pornographic content and impersonation of rivals. “We refuse. Our focus is image enhancement, not manipulation.” Tools parties use to make AI videosCost of a fake Global lessons