Indian election campaigns demonstrate all possible areas of neural network application: from deepfakes to the creation of campaign materials in local dialects.
India has become one of the first countries where artificial intelligence in politics has gone from a technological novelty to a fully-fledged tool of election campaigning. A huge electorate, dozens of languages and dialects, and the major role of social networks and messengers have made the country an ideal testing ground for experimenting with neural networks.
Already during the 2024 parliamentary elections, neural networks were used not only for campaigning, but also to create deepfakes, fabricated statements, and manipulative videos. And just a year later, during regional elections in one of India’s poorest states, Bihar, the use of AI became even more widespread, cheaper, and technologically sophisticated.
The 2024 Parliamentary Elections
In the spring of 2024, India held an election campaign for parliament. By that time, the technology had developed sufficiently for neural networks to generate realistic-looking photographs and videos, and political forces made full use of these tools in their struggle for power.
Numerous fake videos appeared online. For example, the leader of the opposition Indian National Congress (INC), Rahul Gandhi, appeared in two different recordings.
In one, a real video of the politician submitting documents to register as a candidate was overlaid with an AI-generated statement by Gandhi in which he allegedly announced that he was leaving the party.
In the second case, an audio recording that appeared during the voting process had Gandhi supposedly swearing allegiance to the leader of the ruling party, Narendra Modi.
Audio tracks were also replaced in videos featuring popular actors. Unknown individuals added audio criticizing the ruling party and calling on people to vote for the INC to video recordings of actors Aamir Khan and Ranveer Singh. These videos are notable because prominent INC members also participated in their dissemination.
The vice president of another opposition party, Aam Aadmi Party (AAP) was also caught distributing a fake. Through his account, an AI-generated audio recording was circulated in which well-known television host Sudhir Chaudhary allegedly conducted an exit poll showing a victory for an AAP candidate. In reality, no such exit poll had taken place.
New Level of Fakes
AI-generated videos were also used to create musical memes. For example, a video went viral showing India’s prime minister and leader of the Bharatiya Janata Party (BJP), Narendra Modi, performing an energetic dance. The politician took the publication humorously, as it worked to his advantage.
Another politician, West Bengal leader and opposition politician Mamata Banerjee, was far less enthusiastic about an AI video featuring her. In the video, she allegedly sings, “Every child in India will say Jai Shri Ram” Jai Shri Ram literally means “Glory to Lord Rama,” but the BJP made this phrase a key slogan for its supporters in the region. The creator of the video effectively forced the BJP’s main opponent in West Bengal to perform a slogan symbolically associated with the BJP, around which she and the party had been engaged in a long-running conflict.
Another ethically highly questionable format was also actively used during the campaign. Parties used the voices and images of deceased politicians who called on their supporters to vote for the parties they had represented during their lifetimes.
After the elections, the BOOM portal conducted a count and determined that, among 258 cases of the dissemination of false information during the election campaign, only 12 involved AI-generated content. For the whole of 2024, the portal counted 108 such cases in elections at various levels.
The 2025 Elections in Bihar
However, neural networks continued to develop, and in 2025 their sphere of application expanded during regional elections in the Indian state of Bihar.
Marketing companies hired by political parties made extensive use of the new tools: using the most advanced neural networks, ChatGPT and Claude, teams created campaign texts in local dialects for specific groups of voters, and then used voice and video generators to provide voiceovers and create videos. The materials were then posted on social networks as well as in WhatsApp and Telegram group chats, providing broad reach among the state’s nearly 75 million voters.
“The voice cloning in local dialects was the most demanded product and a real game changer,” said Prasad, CEO of Lemon Dot Media, in a comment to the Rest of World portal. He noted that the technology helped candidates convey their positions to voters in the most remote corners of the state.
Prasad noted that the use of AI in elections in one of India’s poorest and most densely populated states demonstrated how effective this technology can be for reaching a larger number of voters through hyperlocal content and fewer resources.
“We were able to save a lot of money by investing in AI tools rather than hiring a lot of people,” Prasad said. He pointed out that subscriptions to digital services cost him only about $1,500 per month, which was far less than the amount that would have had to be paid to actual employees.
AI had been used in election campaigns in India before, but what set the Bihar election apart is the sheer ubiquity of AI in every aspect of campaigning, Amitabh Kumar, an adviser at digital rights organization Social & Media Matters. As an example, he cited situations in which AI was used during candidates’ speeches to translate their remarks from English and Hindi into local state dialects.
Another area of AI application was chatbots that answered voters’ questions in the dialect they were accustomed to. Moreover, they continued operating even during the mandatory 48-hour “silence period” before elections, when broadcasting information related to the election campaign is prohibited, because the law does not regulate the operation of chatbots.
Fakes in Bihar
At the same time, state police opened around a dozen criminal cases over the illegal use of AI-generated content. However, apparently, most such cases went unnoticed. Before the elections, generated videos circulated showing politicians, journalists, and celebrities supporting or discrediting candidates, as well as promising free gifts.
A special team of 15 police officers was responsible for identifying fake content, but tracking and detecting content created with AI is a very difficult task, said senior police officer Manavjit Singh Dhillon.
Attempts to Establish Rules for the Use of AI
Obviously, the scale of AI use in election campaigns will only continue to grow, and authorities are therefore attempting to regulate its use. For now, however, this sphere is weakly regulated, leaving large parties with resources for experimentation and adaptation to new technologies at an advantage. In particular, India’s ruling Bharatiya Janata Party won the elections in Bihar.
Rest of World quoted defeated candidate Divya Gautam of the regional Communist Party of India (Marxist-Leninist) Liberation, who described the victory as proof that AI is a trump card for larger and wealthier parties. “It is really not for candidates like me who do not have money to spend on tech and training,” she said.
However, the authorities are attempting to regulate the sphere. In October 2025, a draft law was published aimed at restricting the dissemination of synthetically generated information. In April 2026, India’s Election Commission published rules regulating the use of AI-generated content as part of election campaigns.
Specifically, it requires political parties to label content generated by neural networks, while social networks and media platforms must remove AI content that violates the law or contains disinformation within three hours of being notified. It was reported that more than 10,000 posts violating the new rules had already been removed within a month.
However, it is obvious that this process cannot be stopped through ordinary bans, and each subsequent election will produce increasingly sophisticated examples of AI use, including for manipulating voters. Civil society can counter this process only through growing critical thinking and the ability to identify inconsistencies.
