Report on AI-weaponised disinformation and institutional failure in the 2026 Assam election

By A Representative 

The Diaspora in Action for Human Rights and Democracy (DAHRD) has released a comprehensive monitoring report documenting what it describes as the most sophisticated AI-enabled disinformation and exclusion operation ever recorded in an Indian election. Ahead of the April 9, 2026, Assam Assembly elections, the monitoring body identified a coordinated architecture of digital propaganda, legislative action, and administrative machinery targeting the state’s Bengali-speaking Muslim community. 

The report details a six-tier content ecosystem that produced 432 confirmed AI-generated posts, accumulating over 45.4 million views. This industrial-scale operation was designed to manufacture an altered reality in which an entire community is simultaneously dehumanized and disenfranchised.

A key finding of the investigation is the existence of a propaganda-to-policy pipeline, where synthetic content establishing narratives such as "Land Jihad" translated directly into enacted legislation, specifically a February 2026 property restriction law prohibiting land sales to Muslims in designated areas.  

​The report highlights a documented vocabulary shift by Chief Minister Himanta Biswa Sarma, who allegedly rotated terms from "Miya" to "Bangladeshi" to "Encroachment" to evade legal and constitutional liability while maintaining identical targeting. Sarma reportedly confirmed this shift in a March 12 interview, stating that earlier content was "constitutionally and legally wrong" because it had not used the word "Bangladeshi". 

Beyond rhetoric, the report monitors a systematic "exclusion architecture" consisting of four simultaneous operations: dehumanization through AI, the deletion of approximately 2.43 lakh names from voter rolls via the Special Intensive Revision process, the physical removal of residents through forced evictions celebrated on social media, and the cultural erasure of figures like the Sufi saint Azan Fakir from Assamese history.  

​High-level political figures were directly implicated in the dissemination of disinformation, according to DAHRD. The report documents 31 confirmed deepfakes targeting opposition candidate Gaurav Gogoi, including a video from a verified Cabinet Minister’s handle that fabricated Gogoi’s identity as a Pakistani agent. Furthermore, the campaign utilized unprecedented gendered disinformation, targeting Gogoi’s wife, Elizabeth Colburn, a private individual, with at least six AI-fabricated intimate and communal scenarios. 

Despite 119 documented breaches of the Model Code of Conduct, including 84 classified as high severity, the report notes a total institutional failure, with the Election Commission of India taking no enforcement action and social media platforms executing zero content takedowns. DAHRD warns that Assam has served as a laboratory for these techniques, noting that a similar administrative architecture has already been used to suspend over 10 million voters in West Bengal. 

The report concludes that without urgent intervention, the widening gap between AI propaganda production and democratic accountability will further compromise the integrity of the 2029 general elections.