The 2024 Indian general election — the world's largest democratic exercise, involving 968 million eligible voters across 543 constituencies over 44 days of polling — was also the world's most AI-contested election to date. Over 50,000 AI-generated deepfakes of political candidates, party leaders, and electoral officials were identified and reported across platforms during the campaign period, according to cybersecurity firm McAfee's post-election analysis. WhatsApp groups — the primary political communication channel for hundreds of millions of Indian voters — were saturated with AI-generated audio clips in regional languages placing fabricated statements in political leaders' mouths. In at least three constituencies, viral deepfakes of local candidates making communally inflammatory statements were identified as factors in the post-poll violence that followed close results.

The Election Commission of India's response — advisories to political parties, takedown requests to platforms, and the deployment of a dedicated misinformation monitoring cell — was the infrastructure of a regulatory body that had not anticipated the scale or sophistication of AI-generated electoral interference. It was not adequate. And the 2024 election was only the beginning.

The Technology Landscape

The tools that produce electoral deepfakes have become democratically — in the worst sense — accessible. In 2020, generating a convincing deepfake video required significant technical expertise, expensive hardware, and hours of processing time. In 2026, several freely available mobile applications can generate a photorealistic deepfake video of any public figure in under three minutes, requiring only a reference photograph and a text script. Audio deepfakes — which are even more difficult to detect than video and which travel further and faster on WhatsApp — can be generated in seconds using voice cloning tools available for free online.

The detection technology has not kept pace. The best deepfake detection models — trained on large datasets of known fakes — achieve accuracy rates of 80–85% on synthetic test data. In real-world electoral conditions, where deepfakes are generated using the latest models specifically designed to evade detection, accuracy rates are significantly lower. Platform-level detection is further complicated by the end-to-end encryption of WhatsApp, which prevents content scanning without compromising user privacy. The technology gap between generation and detection is widening, not narrowing.

The Regulatory Gap

India's current regulatory framework for deepfakes is fragmented across multiple instruments that were not designed for the specific challenge of AI-generated electoral content.

The IT Act 2000, the IT Rules 2021, and the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules 2023 collectively create takedown obligations for platforms hosting deepfake content — but the obligation attaches only after content is reported, the takedown process takes days in a news cycle that moves in hours, and the rules do not distinguish between electoral and non-electoral deepfakes in their urgency requirements. The Election Commission's Model Code of Conduct prohibits electoral misrepresentation but was drafted in an era of pamphlets and cable television, not AI-generated content at scale.

The most significant regulatory gap is the absence of a mandatory disclosure requirement for AI-generated electoral content. Several democracies — including the United States at the state level and the EU under the AI Act — now require that AI-generated content used in electoral advertising be clearly labelled. India has no equivalent requirement. A candidate who uses AI to generate a video of their opponent making statements they never made faces no regulatory consequence beyond the general prohibition on electoral misrepresentation — a prohibition that is difficult to enforce when the content is generated and disseminated in hours and the legal process takes months.

What India Must Build

India's democratic infrastructure needs three specific additions to manage AI-generated electoral interference.

The first is a mandatory AI content disclosure framework for electoral communications — requiring that all AI-generated content used in electoral campaigns, advertising, or social media posts be labelled with a standardised disclosure, with penalties for non-disclosure that are enforceable within the election cycle rather than after it.

The second is a real-time deepfake response mechanism within the Election Commission — a dedicated unit with AI detection capability, platform liaison authority, and the legal power to issue emergency takedown orders for verified electoral deepfakes within hours rather than days. The existing misinformation monitoring cell needs both the technical tools and the legal authority that its current mandate does not provide.

The third is platform accountability for algorithmic amplification — ensuring that social media platforms cannot algorithmically amplify electoral deepfake content even when its origin is obscured. The EU's Digital Services Act's requirement for algorithmic transparency in political content applies a useful model. India's equivalent should be built into the anticipated Digital India Act before the next election cycle.

India is the world's largest democracy. Protecting its electoral integrity from AI-generated interference is not a technology question. It is a constitutional imperative.

The Hind covers policy, power, and strategic affairs from India's perspective. Views expressed are analytical and editorial.