The video posted by a state department of India’s ruling Bharatiya Janata Celebration (BJP) confirmed Assam chief minister Himanta Biswa Sarma capturing a picture of two males in Muslim cranium caps. “Foreigner-free Assam”, learn one caption throughout the video. “Why did you not go to Pakistan?” stated one other.
One of many males within the picture that Sarma was portrayed as capturing was Gaurav Gogoi, a pacesetter of the Indian Nationwide Congress (INC), the BJP’s predominant competitor in Assam for the state’s upcoming legislative elections subsequent month.
Gogoi has said that he’s Hindu however enjoys visiting completely different spiritual websites and observing their norms. He has been photographed sporting conventional Muslim apparel throughout spiritual events equivalent to Eid.
However the picture of him within the video shared by BJP Assam, sporting an informal singlet with a cranium cap, was not a type of events.
Bellingcat has seen a number of dozen movies posted by the BJP that use generative synthetic intelligence (AI) alongside anti-Muslim and anti-Bangladeshi messaging within the border states of Assam and West Bengal in December final yr, forward of legislative elections scheduled in each states for April.
Bellingcat analysed 499 social media posts containing photographs and movies shared on Fb, Instagram and X by the BJP’s official accounts within the two states for this time interval, discovering 194 posts that appeared to satisfy the United Nations’ definition of hate speech: discriminating in opposition to individuals or communities primarily based on inherent traits equivalent to faith and nationwide origin. Of those, 31 (about one in six of the hateful posts) contained the plain use of AI-generated imagery.
Chart: Galen Reich
These look like half of a bigger sample of politicians and events globally utilizing generative AI to amplify hateful or divisive content material, notably forward of main political occasions equivalent to elections.
Forward of the New York Metropolis mayoral race final yr, Andrew Cuomo’s official X account shared, then deleted, an AI-generated video depicting Mamdani consuming rice along with his palms and a Black man in a keffiyeh shoplifting. In Italy, a number of opposition events complained to a communications watchdog after deputy prime minister Matteo Salvini’s League occasion printed a collection of AI-generated pictures depicting males of color attacking girls or law enforcement officials. And within the UK, movies by an AI-generated rapper funded by the far-right Advance UK occasion, with lyrics focusing on Muslims, had been seen hundreds of thousands of occasions.
A Marketing campaign of Hate
Each Assam and West Bengal share a border with Bangladesh. BJP, the world’s largest political occasion, is presently in energy in Assam, the place legislative elections are scheduled on Apr. 9. West Bengal, which fits to the polls on Apr. 23, is ruled by the Trinamool Congress (TMC).
Tensions between India and Bangladesh worsened after former Bangladeshi Prime Minister Sheikh Hasina, who enjoys shut ties with Delhi, was ousted in 2024 and fled to India.
US-based worldwide affairs knowledgeable Mohammed Zeeshan instructed Bellingcat that the “dehumanising and debasing” terminology utilized in India to check with alleged unlawful Bangladeshi immigrants, together with by senior ministers, has precipitated resentment in direction of India in Bangladesh.
“The state of affairs, actually, was so unhealthy that Hasina herself had subtly warned the Modi authorities in public statements that Indian home rhetoric was endangering Bangladeshi Hindus, who bore the brunt of that resentment,” Zeeshan stated.
Zobaida Nasreen, a professor of anthropology at Dhaka College, stated that anti-Muslim rhetoric intensified by BJP leaders reinforces the assumption in Bangladesh that Muslims and Bengalis are being collectively focused in India.
“Viral movies containing this message are inclined to unfold shortly throughout Bangladeshi media and social platforms particularly on Fb, enhancing perceptions of hostility and triggering anti-India sentiment or nationalist backlash,” she added.
In December, the month our dataset was collected, Dipu Das, a Hindu garment employee, was crushed to dying at an anti-India protest in Bangladesh over allegations that he had made derogatory remarks about Islam.
And whereas the administration led by Bangladesh’s newly elected chief Tarique Rahman has sought to reset strained ties, many of the hateful social media posts we noticed posted by the BJP in December attacked Bangladeshi Muslims and/or Bengali-origin Muslims in India, exhibiting how tensions between the 2 nations proceed to affect political messaging in India’s border states.
Bellingcat’s evaluation included a complete of 202 posts by BJP Assam and 297 by BJP’s West Bengal department on their official accounts. We additionally checked out posts shared by BJP’s predominant opponent events – 194 from INC in Assam and 357 from the TMC in West Bengal – throughout the identical time interval in December.
This included all visible social media posts (containing photographs or movies) by every occasion in December, besides those who didn’t seem to comprise any overt political messaging, equivalent to these merely commemorating public holidays. We solely counted every picture or video as soon as, no matter what number of platforms it was shared throughout.
Though the entire main events contesting within the Assam and West Bengal state elections appeared to make use of AI-generated imagery in a few of their posts, there gave the impression to be a very excessive focus of hateful messaging within the ones posted by the BJP’s accounts.
In Assam, we recognized 28 posts by BJP utilizing apparently AI-generated imagery, of which 24 carried hateful messaging. Of the 194 INC posts we checked out from December, 41 appeared to characteristic AI-generated imagery, however none of those appeared to hold hateful messaging.
In West Bengal, we discovered 14 BJP posts that contained clear indicators of AI-generated imagery, seven of which had been hateful. We additionally recognized 15 posts by the incumbent TMC that appeared to characteristic AI imagery, however none of those appeared to satisfy the definition of hate speech.
Assist Bellingcat
Your donations straight contribute to our means to publish groundbreaking investigations and uncover wrongdoing all over the world.
When contacted for remark, BJP Assam spokesperson Rupam Goswami didn’t straight reply to questions on the occasion’s normal use of AI however stated they didn’t submit any AI-generated photographs of Gogoi. “BJP doesn’t stoop so low,” he instructed Bellingcat.
As for the “level clean” capturing video, Goswami initially stated the individual accountable had been punished and faraway from the occasion. Nevertheless, when requested about Sarma saying that he would re-post the video with these he was depicted capturing labelled as “Bangladeshis”, Goswami stated, “[Bangladeshis] must be fully suppressed.”
BJP West Bengal didn’t reply to a number of requests for remark by Bellingcat by way of telephone and e mail.
It is very important observe that as generative AI know-how improves, it may be more and more troublesome to detect AI-generated imagery. Our guide depend of AI-generated imagery solely included posts that had apparent indicators of generative AI equivalent to unnaturally clean textures and a number of folks with the identical faces. It’s subsequently attainable that there have been different pictures in our dataset the place generative AI was used extra subtly.
Nevertheless, Joyojeet Pal, Professor of Data on the College of Michigan, instructed Bellingcat that the standard of those visuals, or whether or not they regarded actual, was not the precedence.
“What politicians in India have understood is that the sociocultural drivers of misinformation are most vital for elections, so that they harp on about issues to the extent that they’ve began to not care about kind over substance. It seems unhealthy? It doesn’t matter,” he stated.
Extra vital to voters, based on Pal, was whether or not they already believed within the narrative contained within the movies, which generative AI may assist create extra shortly: “AI helps cement polarised opinions by providing you with the type of content material you’ve gotten already determined you wish to interact with.”
When requested about INC’s use of AI, occasion spokesperson Aman Wadud stated that it was apparent that a number of the movies they posted had been made with AI and that there was no intention to mislead.
“AI might be each harmful and artistic. We’re utilizing it in a artistic method, we aren’t utilizing it in a harmful method. We don’t violate folks’s dignity, we don’t falsely accuse folks,” he stated.
TMC didn’t reply to Bellingcat’s a number of requests for remark by way of telephone and e mail by publication time.
Portraying Bengali Muslims as ‘Foreigners’
The biggest class of hateful messaging Bellingcat noticed within the BJP’s posts focused Bangladeshi or Bengali-origin Muslims, referring to them as “infiltrators” or “foreigners”. We counted 66 such posts by the BJP’s Assam and West Bengal branches from December, of which eight appeared to comprise apparent AI-generated imagery.
Bengali-origin Muslims are usually stereotyped as “unlawful immigrants” within the state, though members of the neighborhood have lived in India because the late 1800s.
Final yr, the BJP deported hundreds of alleged undocumented migrants – reportedly together with Indian Muslim residents – to Bangladesh. Human rights teams have referred to as the deportations illegal and discriminatory, in addition to missing in due course of.
One video referencing this theme reveals AI-generated visuals of protests in opposition to “unlawful infiltration” in Assam, with the caption urging folks to “get up” or the nation would “flip into Bangladesh”.
A special one makes use of actual footage from previous violence in Assam blended in with pictures of Muslim males. A tune enjoying within the background accuses them of taking up “Assamese land” and reveals AI pictures of “Assamese” folks, i.e. these not in stereotypical Muslim clothes, crying.
Each movies use spiritual markers to attract a distinction between “infiltrators” – males in cranium caps or lungis related to Bengal-origin Muslims – and “residents” in non-Muslim apparel.
Clothes is usually utilized by the Hindu far-right as a visible shorthand for id and a deepening spiritual divide. In 2019, Prime Minister Narendra Modi stated of protests in opposition to a controversial citizenship legislation that these accountable for violence could possibly be “recognized by their garments”.
Within the hateful posts seen by Bellingcat, each actual and AI-generated pictures of opposition figures – notably Gogoi – had been proven alongside messaging that urged that they supported “foreigners” or “infiltrators”.
The Heart for the Examine of Organized Hate (CSOH) additionally famous, in a 2025 report on AI-generated imagery and Islomophobia in India, that Hindu far-right politicians and media shops have invoked and strengthened the trope of Muslims as “infiltrators” for years.
“AI-generated pictures on these themes reinforce associations between Muslim id and illegality, reinforcing xenophobic and Islamophobic stereotypes. In doing so, they play a robust position in justifying exclusionary insurance policies and normalising discrimination in opposition to Muslims,” the report stated.
‘Save Hindus’
Zenith Khan, a knowledge analyst who labored on the CSOH report, famous that AI-generated propaganda was usually tightly knit with present political moments, and its impression relied on “timing it proper” particularly when “individuals are emotionally charged”.
The violence in opposition to the minority Hindu neighborhood in Bangladesh has been utilized by the BJP to lift issues over the security of Hindus in India.
Days after Das’ lynching, the Assam state department of BJP posted a video with a picture of his face – besides that it was manipulated with AI to point out tears streaming from his eyes. “Save Hindus”, stated the textual content accompanying the video.
Posts by BJP’s West Bengal unit additionally appeared to border Muslims as criminals or threats. A video, styled after the TV present “Stranger Issues”, raised alarms over an “the wrong way up” model of the state underneath the present authorities.
A person is depicted being chased by males in cranium caps. Arrows label them as “Ralib,” “Galib,” and “Chalib” – a play on Muslim names ending in “-lib” – in case the cranium caps left any ambiguity about their Muslim portrayal.
INC filed a police grievance in September final yr in opposition to the BJP for sharing AI movies focusing on Gogoi and the Muslim neighborhood, in addition to one other grievance in relation to the video of Sarma portrayed as capturing two males “level clean” in February.
INC Assam spokesperson Wadud stated that no motion had been taken on the occasion’s police complaints so far as he knew.
Disinformation researcher Bharat Nayak instructed Bellingcat that it has all the time been tech platforms’ duty to manage new sorts of content material.
“The purpose submit can’t shift. This has all the time been a tech downside,” he stated.
When this duty is shrugged off, Nayak added, the result’s an absence of accountability. “When you’re utilizing outdated movies from different nations as new, you should have folks countering you. However AI-generated movies might be shared with out context simply to unfold hate – like exhibiting folks in cranium caps – and the ‘when, the place, how’ questions vanish.”
Each Meta – which owns Fb and Instagram – and X have insurance policies in opposition to hateful conduct.
Meta additionally introduced in 2024 that it could begin including “AI data” labels to extra content material detected as AI-generated, whereas some X customers noticed an analogous characteristic launched on the platform final month. Solely 5 of INC’s AI visuals that we recognized – and none of these by TMC or the BJP – had a disclaimer that stated “AI-generated”.
Bellingcat reached out to Meta and X for touch upon whether or not the posts we recognized breached their phrases of use concerning hateful conduct or labelling AI-generated posts. A Meta spokesperson stated they had been reviewing the flagged content material and “will take acceptable motion on any violations of our insurance policies”. As of publication, X had not responded.
Kalim Ahmed from Bellingcat’s Discord Neighborhood contributed analysis to this piece.
Bellingcat is a non-profit and the flexibility to hold out our work depends on the type assist of particular person donors. If you need to assist our work, you are able to do so right here. You can too subscribe to our Patreon channel right here. Subscribe to our E-newsletter and comply with us on Bluesky right here, Instagram right here, Reddit right here and YouTube right here.
