The Dark Side of Generative AI: Deepfakes, Elections, and India’s Regulatory Vacuum

The same generative AI capabilities that allow a small business to create marketing content or a student to get tutoring assistance also allow the creation of convincingly fabricated video and audio content depicting real public figures saying things they never said. India, with its large population, multiple languages, and politically intense electoral cycles, presents a particularly consequential testing ground for this technology’s darker applications — and the legal and regulatory framework has not kept pace with the speed at which the underlying tools have become accessible.

The Technology Has Outrun the Law

Creating a passably convincing deepfake video or voice clone, which required substantial technical expertise and computing resources only a few years ago, is now achievable with consumer-accessible tools and minimal technical skill. During recent election cycles, multiple instances of AI-generated or AI-manipulated audio and video content depicting political figures have circulated widely on social media and messaging platforms before being identified and debunked — a debunking that, by the nature of how misinformation spreads, typically reaches a fraction of the audience that the original content reached.

The Specific Indian Vulnerability

India’s information ecosystem has structural characteristics that make it particularly susceptible to AI-generated misinformation at scale. A very large share of political content consumption and sharing occurs on closed messaging platforms, where content moves through private networks largely outside the reach of public fact-checking organisations and platform content moderation systems designed primarily for public social media feeds. Combined with India’s linguistic diversity, which strains the capacity of both AI detection tools and human fact-checking organisations to operate at scale across every language in which misinformation might circulate, the conditions for deepfake-driven electoral disruption are unusually favourable.

A law designed for an era when fabricating convincing video required a film studio cannot meaningfully govern an era when it requires a laptop and an afternoon. India’s information law has not yet caught up to this gap.

The Regulatory Response So Far

India’s existing IT Rules have been amended to require labelling of AI-generated content on major platforms and to mandate faster takedown timelines for flagged misinformation, and the Election Commission has issued specific advisories to political parties regarding the use of AI-generated content in campaign material. These measures represent meaningful first steps, but they rely heavily on platform self-reporting and labelling compliance, mechanisms that have proven only partially effective at the speed and scale at which viral content spreads on closed messaging platforms in particular.

A more robust framework would likely require dedicated, well-resourced rapid-response fact-checking capacity specifically built for closed-platform misinformation, clearer statutory liability for platforms that fail to act on flagged deepfake content within defined timelines, and public AI detection tools made freely available to journalists and citizens rather than relegated to specialist research labs. Until these capabilities are built at the scale India’s information ecosystem requires, each election cycle is likely to serve as a live, uncontrolled experiment in AI-driven misinformation’s electoral impact — an experiment India can ill afford to keep running without better safeguards in place.