Author: Bhagyashree Dhingra
In April and May 2026, five Indian states and a Union Territory – Assam, Kerala, Tamil Nadu, West Bengal and Puducherry – went to the polls in one of the largest democratic exercises the country has seen this decade, covering over 17 crore voters and more than 800 assembly seats.
What made this election cycle different from anything before it was not just the size of the electorate, but the presence of a new and unpredictable participant in the campaign: artificial intelligence. From AI-generated videos of deceased political icons “addressing” rallies to cloned voices of leaders making promises they never made, the 2026 elections marked the moment India had to confront the deepfake problem head-on, not as a hypothetical future risk, but as a live election-season reality.
The Problem: When Seeing Is No Longer Believing
A deepfake is a piece of synthetic media – video, audio or image – created using artificial intelligence to make it appear that a real person said or did something they never actually said or did. What was once an expensive, niche technology requiring specialist skills is now available through cheap, user-friendly apps that anyone with a smartphone can use. In the run-up to the 2026 state elections, reports surfaced of AI-generated videos featuring deceased leaders such as M. Karunanidhi and M.G. Ramachandran in Tamil Nadu, voice clones of senior leaders making statements during the campaign, and hyper-targeted memes designed to mislead specific voter groups based on language, religion or caste.
The danger here is not merely that false information is being spread – misinformation has always existed in politics. The danger is the realism. When a voter sees a video of a familiar face speaking in a familiar voice, the natural instinct is to believe it. Deepfakes exploit this trust, and in a country where a significant share of the electorate consumes political content primarily through WhatsApp forwards and short videos, the potential for large-scale manipulation is enormous. Unlike a printed pamphlet with a known publisher, a deepfake video can spread
without any clear point of origin, making it extraordinarily difficult to trace, attribute and counter before it influences voters.
The Legal Response: From Silence to a Regulatory Framework
For years, India’s legal framework dealing with deepfakes was indirect at best. The Information Technology Act, 2000 and the IT Rules, 2021 placed obligations on social media intermediaries to remove unlawful content within a defined period, and provisions such as Section 66D of the
IT Act dealt with cheating by personation using a computer resource. The Indian Penal Code’s provisions on defamation, forgery and cheating could technically be stretched to cover some deepfake-related harms. However, none of these provisions were designed with AI-generated synthetic media in mind, and enforcement was slow, reactive and inconsistent.
The lead-up to the 2026 elections changed this. The Ministry of Electronics and Information Technology introduced amendments to the IT Rules that, for the first time, expressly recognise “synthetically generated information” (SGI) and deepfakes as a distinct regulatory category. The amended rules identify a category of “prohibited SGI” that includes deceptive political or electoral content, impersonation or fraud-oriented deepfakes, and forged material capable of misleading the public or authorities. Crucially, the rules also compress the takedown timeline for such content: where the earlier regime allowed intermediaries up to 36 hours to act on a government or court order, the new framework requires action within three hours for certain categories of AI-generated content.
Alongside this, the Election Commission of India rolled out its first comprehensive set of guidelines specifically addressing AI use in campaigning for the 2026 state elections. These guidelines mandate visible disclosure labels – such as an on-screen “SYNTHETIC” marker covering a portion of the screen – for AI-generated videos of political figures, and audio disclaimers at the start of cloned-voice content. They also set up a faster complaint-to-takedown pipeline, allowing citizens to flag suspected deepfakes through the existing cVIGIL app, which then triggers metadata verification, a platform removal notice and, where necessary, a First Information Report.
Constitutional and Practical Tensions
While the intent behind these measures is undeniably welcome, they raise genuine constitutional questions, particularly around the freedom of speech and expression guaranteed under Article 19(1)(a) of the Constitution. A three-hour takedown window, while useful for containing viral falsehoods quickly, leaves very little room for intermediaries to verify whether content actually falls within the “prohibited” category before acting. Civil society groups and digital rights organisations have already raised concerns that such compressed timelines could lead to over-removal of legitimate political satire, commentary or criticism out of an abundance of caution, effectively creating a chilling effect on speech that has nothing to do with deception.
There is also the question of enforcement against political parties themselves. Much of the deepfake content circulating during the 2026 campaign was not the work of anonymous trolls but appeared to originate from, or be amplified by, party-affiliated social media networks. A regulatory framework built primarily around obligations for intermediaries does relatively little if the original creators of misleading content are political actors who have both the motive and the resources to keep producing new variants faster than they can be taken down. The labelling requirements imposed by the Election Commission are a step in the right direction, but their effectiveness depends entirely on consistent monitoring and a willingness to act against powerful political players, not just fringe accounts.
Why This Matters Beyond the 2026 Elections
The significance of this moment extends well beyond the five states that voted in April and May 2026. India is set to hold a continuous cycle of state elections over the coming years, and the precedent set now – in terms of how deepfakes are defined, how quickly they must be removed, and how political parties are held accountable – will shape the information environment for every election that follows. If the current framework proves workable, it could become a template not just for India but for other large democracies grappling with the same problem. If it proves either too weak to stop coordinated disinformation campaigns or too broad in a way that suppresses legitimate speech, India will need to revisit it quickly, because the technology enabling deepfakes is only becoming cheaper, faster and more convincing with each passing month.
Conclusion
The 2026 state elections will likely be remembered as the point at which India’s election law formally acknowledged the deepfake era. The amended IT Rules and the Election Commission’s AI guidelines represent a meaningful first attempt to bring synthetic media within a clear legal framework, balancing the need for rapid response against the risk of misuse. But law alone cannot solve what is fundamentally a problem of trust and media literacy. As voters, the most effective safeguard remains a healthy scepticism towards anything that seems designed to provoke an immediate emotional reaction – a reminder that, in this new digital landscape, the old advice to “verify before you believe” has never been more legally and politically relevant.
