AI-Driven Disinformation Reshaping India's Elections
Artificial intelligence is transforming India's electoral landscape through the rapid spread of disinformation, from fabricated news to cloned candidate voices. The proliferation of deepfakes is blurring lines between legitimate campaign strategy and malicious interference. This is creating urgent demand for regulatory and technical solutions like rapid takedown mandates and content provenance labeling.
- During the 2024 general election, political parties used AI to resurrect deceased political figures like M. Karunanidhi and J. Jayalalithaa to deliver speeches and endorsements, playing on voters' emotional connections. Other deepfakes featured Bollywood actors Ranveer Singh and Aamir Khan criticizing the prime minister, prompting them to file police complaints. - In response to the growing use of deepfakes, the Election Commission of India (ECI) has mandated that political parties must clearly label all AI-generated content. Parties are also required to take down any deepfake audio or videos from social media platforms within three hours of their discovery. - Voice cloning was widely used to create personalized messages from candidates, with over 50 million such calls made in the two months before the April 2024 elections began. This tactic was seen as more authentic than video deepfakes and was used by parties of all sizes, including one to broadcast messages from a party leader who was in jail. - India's Ministry of Information Technology amended the IT Rules, shortening the timeline for social media platforms to remove unlawful content from 36 hours to just three hours after receiving a government or court order. For urgent situations, like non-consensual deepfake imagery, the removal time is two hours. - While only 2% of misinformation reviewed by Google's Project Shakti was AI-generated, parties were projected to spend around $50 million on authorized AI content during the 2024 elections. This included using AI for positive outreach, such as Prime Minister Narendra Modi using the Bhashini translation tool to deliver speeches in multiple Indian languages simultaneously. - A pan-India consortium of news publishers and fact-checkers called Shakti was formed to create a common repository to aid in the early detection of online misinformation, including deepfakes. India now has 17 International Fact-Checking Network (IFCN)-certified fact-checking organizations, the most of any country. - The first notable use of deepfakes in Indian politics occurred before the 2020 Delhi Assembly Election, when two videos of a party president were altered to show him speaking in different languages and sent to 15 million voters. - An operation first uncovered by OpenAI, codenamed "Zero Zeno," was identified as an AI-generated misinformation network targeting the Indian elections. The campaign created and disseminated content critical of the ruling Bharatiya Janata Party (BJP) while promoting the opposition Congress party across various social media platforms.
