The Assam Assembly election campaign witnessed what an advocacy group described as the “first large-scale artificial intelligence-driven disinformation operation” in a state election in India, according to the report released by the Diaspora in Action for Human Rights and Democracy.
The report was released just days before the Assam Assembly election on Thursday whose results will be announced on May 4.
The organisation further reported that the disinformation campaign was a part of broader strategy in which Muslims were “simultaneously dehumanised, disenfranchised, displaced and erased from cultural memory.”
The report highlighted that it identified 31 confirmed deepfakes targeting Assam Congress Chief Gaurav Gogoi by “fabricating his identity as a Pakistani agent and Muslim sympathiser”.
“Thirty-one confirmed deepfakes targeted Congress CM candidate Gaurav Gogoi, fabricating his identity as a Pakistani agent and Muslim sympathiser, distributed through official BJP accounts and a verified Cabinet Minister’s handle (Pijush Hazarika; 97,700 views). Six AI-fabricated videos targeted Gogoi’s wife Elizabeth Colburn — a private individual holding no public office — with intimate and communal scenarios. No comparable deepfake campaign targeting a political spouse has been documented in any previous Indian election,” the report said.
“119 Documented Breaches”
Additionally, it highlighted that “119 documented breaches” of the model code of conduct, and reported that no actions in any such cases were taken by the Election Commission. Furthermore, the report said that the social media platforms also did not take down any such content.
The report said that the model implemented in Assam – voter roll purges, demographic engineering and AI-generated communal content – was being replicated in other parts of the country, including poll-bound West Bengal.
The organisation monitored 273 accounts across Facebook, Instagram, and X, amounting to an audience of 407.4 million people. In the time period of five months starting from November 2025 and April 2026, the organisation examined 2,462 posts, which identified 432 AI generated posts alone, reaching 45.4 million views.
“Even at one percent engagement, a single AI post reaches four million people — over one-sixth of Assam’s registered voters,” the report states.
The report said that a large network of “disinformation architecture” which included synthetic images, deepfake videos, and AI generated communal content was made ahead of the polls.
432 posts of Facebook and Instagram “likey” “very likely” Made using AI
It said 432 posts of Facebook and Instagram which were “likely” or “very likely” made using AI, getting around 45.4 million views and as many as 1 lakh likes. It also noted that one Instagram account, “politooons”, alone generated about 40.2 million views through 102 AI-made posts, making up 88% of the total views of such content.
“A total of 432 posts across Facebook and Instagram were classified as Very Likely or Likely AI-generated using a 13-signal detection rubric, generating 45.4 million views and over 100,000 likes. The operation was industrialised, not improvised: a six-tier content ecosystem produced synthetic images, deepfake videos, and AI-generated communal content at volume. A single Instagram account — politooons — generated 40.2 million views across 102 AI posts, accounting for 88 per cent of all AI content views. The operation was front-loaded: 70 AI posts in January, 58 in February, 18 in March — building the narrative frame months before the Model Code of Conduct took effect,” the report said.
The report also highlighted that the Assam BJP earlier posted a video which was later removed, demonstrating the Chief Minister Biswa Sarma symbolically firing at images of two muslim men. However, the clip was deleted ahead of the social media criticism.
“Four Operations Against Assam’s Muslim During Single Active Elections“
Dehumanisation
The report revealed that 432 AI posts and 18 verified statements by CM Himanta Biswa Sarma showed muslims as “existential threat”, using slur “Miya”, a derogatory word used to refer undocumented immigrants, appeared in 76 per cent of all Assamese-language posts monitored.
Electoral Exclusion
It further highlighted that the CM publicly declared “4 to 5 lakh Miya votes will be deleted” through the Special Revision process. Subsequently, as reported, the 2.43 lakh names were omitted from Assam’s voter rolls.
Physical removal
The report said that over three years, 68 social media posts promoted forced evictions as a government success. Posts by Chief Minister Himanta Biswa Sarma, where people were labelled as “Bangladeshi”, received an average of 5,818 likes — about 4.5 times higher than the overall average.
The report also criticised regulatory bodies, claiming: “The Election Commission of India took no enforcement action against 119 documented MCC breaches.” It also alleges that “social media platforms executed zero content takedowns” and that Meta’s AI labelling policy “produced zero labels on 172 AI-flagged posts.”
The described “on-record admission of deliberate legal evasion” by the Chief Minister and quoted an interview, “We had not added the word Bangladeshi — it was constitutionally and legally wrong. But we will correct it and post it again.”
“The 8 February Flood: 14 Posts in One Week“
The reported noted that Ahead of a BJP press conference, coordinated AI content had already been made rounds on internet, which included “Paaijaan” themed deepfakes in a week and the hashtags #PakAgentGaurav which was used 28 times in a day, highlighting a structured disinformation strategy.
“The Propaganda Became Law“
It also called Assam as a “laboratory,” that would spread similar strategies in the nation. “The structural gap between AI propaganda production capacity and democratic accountability mechanisms is widening with each election,” it said.
“For the first time… AI-generated content translated directly into enacted legislation within a single election cycle… The propaganda became law,” the report claims.
Read Full Report Here.
