AI and India’s General Elections

A video of Bollywood actor Aamir Khan mocking India’s ruling Bharatiya Janata Party (BJP) for not fulfilling a decade-old promise to deposit 1.5 million Indian rupees (or US$ 18,000) into the bank account of every Indian citizen has gone viral this election season. The voice in the video closely resembles Khan’s voice, and the video ends with a call to voters to support the opposition Congress party. The following day, Khan dismissed the video as “fake and completely untrue.”

It turns out that the video was manipulated with artificial intelligence (AI) using deepfake audio technology.

It appears that the makers of the video were looking to cash in on Khan’s popularity to attract votes and turned to AI technology to fake his voice. This is just one example of how AI is being used in the ongoing 18th general elections in India.

This is not the first time that the Indian political world witnessed the deployment of AI for the communication of messages. Such technology has been used for over a decade and the BJP has been at the forefront of leveraging digital tools.

During the 2014 Indian general elections, the BJP employed AI-driven tools to target voters with personalized messages, setting a precedent for the integration of technology into political strategies.

“Since then, AI has made deep inroads, influencing how campaigns are conducted and elections are contested,” observes Patarlapati Nagaraju of Osmania University in India.

India witnessed its first known use of a deep fake video, during the Delhi Legislative Assembly elections in February 2020. Two manipulated videos featured Manoj Tiwari, the then-president of the BJP’s Delhi unit. In one he is seen talking in English and in another, in Haryanvi, to connect with a specific demographic of voters. Although the BJP sought to justify its action by saying that it had used AI technology positively to communicate a BJP leader’s message and not to deride a rival, the videos triggered heated discussion as they raised concern about the potential for use and misuse of AI for misinformation and manipulation in political campaigns.

The value of AI technology in communication and reaching out to diverse populations was underscored during the Kashi Tamil Sangamam event in Varanasi. At the inauguration on December 17, 2023, Prime Minister Narendra Modi took the stage. As he delivered the inaugural speech, his voice, familiar to many, began to resonate in the crowd.

Even as he spoke, an AI-based translation tool named “Bhashini” was silently at work. For the first time, Modi’s speech in Hindi was being translated in real-time into Tamil. This remarkable feat not only bridged the linguistic divide, but also marked a leap forward in India’s technological prowess, showcasing the potential of AI in fostering cross-cultural communication.

The translated speech was then broadcasted across social media platforms, reaching thousands of non-Hindi-speaking people who had never before heard the prime minister speak their language.

That was just the beginning. Since then, not only in Tamil, Modi’s speeches are being translated into Kannada, Bengali, Telugu, Punjabi, Marathi, Odia, and Malayalam using AI.

The use of Bhashini at the Varanasi event was a forerunner to the use of the technology by political leaders to make their messages accessible to a diverse Indian audience, breaking down language barriers and ensuring inclusivity in communication. It evoked mixed reactions.

While many applauded the BJP’s efforts to connect with the people of different regions, others raised concerns about the authenticity of the translations and the potential for misinterpretation.

If the 2019 Indian general election earned the name “social media election” due to the extensive use of social media platforms by all parties for election campaigning, the ongoing general election in India is witnessing the emergence of  AI as a pivotal tool in political strategies.

Spanning over six weeks and seven phases, the Indian general election will see the largest number of voters in the world coming out to exercise their franchise. Adding to the logistical challenges of conducting an election for such a massive electorate is the expenditure involved. In the 2019 election, an estimated  7 billion was spent by the Election Commission, political parties, and candidates. The 2024 election is expected to be even more expensive.

“The massive scale-up in content production and the shift towards digital dissemination highlight a transformative change in how political messages are conveyed, particularly with the rapid increase in political content consumption among rural households,” Vinay Deshpande, chief product officer at Rajneethi, one of India’s leading political management consultancy organizations told The Diplomat.

“This shift underscores the importance of digital narratives over traditional physical rallies, marking a new era in electoral strategies,” he added.

AI is expected to contribute significantly to streamlining the election process and making it more efficient.

According to Sagar Vishnoi, an Indian political campaigner and communication specialist, “One of the primary benefits of integrating AI into election campaigns” is that it will “significantly reduce costs, save time, and enhance overall efficiency and productivity.” He cited the example of AI-powered calls. “Traditionally, political parties relied on call centers staffed by workers and volunteers, or utilized IVR [interactive voice response] systems for outreach.” With AI, “these costs can be reduced by up to 50 times, offering a more cost-effective solution for communication and engagement with voters,” he told The Diplomat. With the integration of AI, political parties and governments will be able to “drastically accelerate” the “speed of conducting surveys,”  Vishnoi said, adding that “what traditionally took months to complete can now be accomplished within a single day.”

According to Deshpande, the use of AI in Indian politics “has brought about both savings and increased expenses. It’s facilitated a significant surge in digital content production, up to twentyfold, without expanding the workforce. However, this efficiency hasn’t translated into reduced costs. Rather, there’s a growing investment in technology to meet the demand for more advanced digital content. Thus, while AI streamlines certain processes, it also drives up tech expenses, consuming a larger portion of parties’ budgets.”

When utilized appropriately, AI has the potential to swiftly gather feedback from a vast number of people, ranging from millions to tens of millions, and display real-time results on a dashboard for analytical purposes. This efficiency extends beyond surveys; AI can also be employed for sentiment analysis, recruitment and selection processes, and aiding political leaders in various capacities.

In India, a major concern about AI relates to the threat posed by deep fake videos. During the Telangana assembly elections last year, a video purportedly featuring the then-Chief Minister K Chandrasekhar Rao and other politicians of the Bharat Rashtra Samithi urging voters to support the rival Congress party. The BRS accused the Congress of using deepfake technology and AI to create and circulate “fake audio and video” content against the BRS.

In February this year, the Congress party further fueled the discourse on AI by posting a video parodying Modi on Instagram. The snippet, extracted from a Hindi music album titled ‘Chor‘ (thief), lampoons Modi for facilitating the transfer of resources to business tycoons. An AI meme uses Modi’s voice and face clones to underscore opposition criticism of the prime minister’s patronage of billionaire Indian businessmen.

Even as AI has come in handy for parties to mock their rivals, it is also useful to revive memories of departed leaders. The Dravida Munnetra Kazhagam, which is in power in Tamil Nadu, has resurrected its founder-leader, the late M. Karunanidhi through deep fake technology. Karunanidhi was a reputed orator and his AI avatar delivers speeches praising the leadership of his son and successor K. Stalin, who is the DMK president and current chief minister of Tamil Nadu.

In the context of generative AI emerging as a “remarkable innovation poised to revolutionize our world,”  politics has become one of the areas “profoundly impacted by this technology,” Senthil Nayagam, founder of Muonium, an AI media tech firm known for its deep fake videos, told The Diplomat.

“Deep fake,” he said, “represents a significant advancement powered by AI.”

Nayagam argued that he views deep fake as ethical “until it deviates into creating unrelated content or manipulating public opinion.”

“Yet, the looming threat of misinformation, especially during elections, underscores the need for vigilance,” he said.

Nayagam pointed out that while there’s currently no law to regulate this technology, “it’s essential to seek permission, particularly when representing deceased figures.”

“Politicians’ legacies, often preserved in books, recordings, and other media, are subject to reinterpretation,” Nayagam said. “While recreating such content may not inherently be unlawful, inserting fabricated words risks misleading the voters.”

As discussed above, AI-facilitated real-time translation of speeches facilitates the carrying of messages to voters speaking different languages. But it carries risks too as real-time translation is tricky. A language has many nuances and most AI models, which have been trained abroad, don’t know the Indian cultural context. Even a small mistake in pronunciation could lead to conflict.

As Ankit Lal, an Indian political campaign strategist and consultant, pointed out, “AI-enabled speech translation from one Indian language to another is still far from perfect. The only way to improve it is by using more data, which is a time-consuming exercise.

According to Lal, one way to resolve this problem is “passing the AI-generated content through a human editor, who has a better understanding of the two languages as well as the context, and following that with a ‘peer review’ to take care of the cultural context and sensitivities.”

How AI might spread lies and twist facts is a big worry, Lal said.

If in the past people were “sharing conspiracy theories in text,” now “AI can make those same lies into videos using the faces and voices of famous people, Lal said.

There is concern too over other countries using AI to disrupt the election environment. Recently Microsoft issued a warning regarding China’s possible deployment of AI-generated content to influence public opinion and advance its geopolitical interests. This content, comprising memes, videos, and audio, could be disseminated to sway election outcomes, although the immediate impact might be limited.

“India doesn’t have a good way to track AI-made content. So, stopping sneaky interference is hard,” Lal said, adding that big platforms like Meta, Google, and X are better at handling this.

India’s election authorities should talk to these platforms and set up a system to protect the elections from bad AI content,” he said.

So, the integration of AI technology into India’s political landscape marks a significant shift in how campaigns are conducted and messages are disseminated. From real-time speech translation to deepfake videos, AI presents both opportunities and challenges for political actors. While it facilitates cross-cultural communication and enhances campaign efficiency, concerns regarding authenticity, misinformation, and foreign interference loom large.