Bangalore, India – In early March, a realistic AI-generated image of Indian Prime Minister Narendra Modi dressed as Bhishma Pitama from the ancient Hindu epic Mahabharata was promoted as a political ad on Instagram.
With long wavy gray hair, a sun-shaped mark on his forehead, and wearing body armor, the image depicts what Modi fans see as his role in today's world: fighting foreign threats. It depicts the reincarnation of Bhishma, the supreme commander.
The Instagram image, created by right-wing page Hokage Modi Sama and first posted in 2023, was promoted as a political ad for two days in March and received more than 35,000 impressions.
Al Jazeera examined data from the meta-ad library of political ads in India over the past three months and found that between February 27 and March 21, Hokage Modi Sama generated around 50 AI-generated images of Modi. It has become clear that the company has become a major advertiser in India. His AI-generated Modi image on Instagram.
The Meta Ad Library is a public archive that hosts a collection of political ads run on platforms like Instagram and Facebook.
A common theme in all the images shared by the handle was Modi's reputation as a Hindu leader. A popular AI image through a sponsored post on 'Hokage Modi Sama' features Modi as the reincarnation of the divine son Bhishma, embracing Hindu traditions and wearing a suit of Hindu Rashtra kings on the throne. It garnered millions of likes and views. (Hindu Rashtra is a controversial ideology of Hindu-majority rule in India that seeks to break away from the nation's secular founding principles.)
“Through these images, the attempt appears to be to give Mr. Modi simultaneously sage-like and warrior-like qualities, both of which make us indomitable, invincible, impeccable, and therefore our It creates an aura of a political leader worthy of the unquestioned loyalty of his people,'' Amog Dar Sharma, a lecturer in political communication at the University of Oxford, told Al Jazeera.
However, the posting of AI images of Prime Minister Modi by the most prolific online advertisers suggests that the manipulated images could be used for propaganda purposes among voters who do not fully understand the photo or the extent of alteration that has been made to it. It also illustrates the challenges of enforcing AI-related rules on social media, amid concerns that AI-related rules could be used on social media. The meme may have passed.
Meta is aware of this use and will begin disclosing the use of political ads on Instagram and Facebook created using artificial intelligence (AI) starting in January ahead of the critical election year of 2024. It was announced that there is a need to do so, and that there is a risk of being banned if the information is not disclosed.
In the 30 days ending March 29, Hokage Modi Sama spent 537,799 Indian rupees ($6,500) posting 363 political posts, including images and videos, on his Instagram page, according to data from Meta Advertising Library. Promoted content. Our analysis shows that nearly 14% of all his sponsored ads (equivalent to 50 images) were generated by AI.
All Hokage Modi Sama AI ads were sponsored posts on Instagram, and disclosure of the use of AI was done through hashtags such as #aiartwork, #midjourneyart, and #midjourneyai.Midjourney, which the hashtag seems to refer to, is a popular generative artificial intelligence program.
But Mehta told Al Jazeera that hashtags are not acceptable as a disclosure of digitally created or modified advertising. If an advertiser is required to disclose that content is digitally created or modified, Meta will include the ad library and add a “Digitally Created” label near the “Paid Amount” disclaimer. Masu. Such labels currently do not exist in Hokage Modi Sama AI advertisements.
Mehta has asked Al Jazeera whether Prime Minister Modi's graphic ads violate Al Jazeera's policy of depicting things that real people are not actually doing or depicting events that do not occur in real life. did not respond to specific questions.
Instead, Mehta pointed Al Jazeera reporters to a March 2024 blog post about AI disclosure requirements for advertisers and the company's preparations for India's elections.
Prateek Wagre, executive director of the India-based Internet Freedom Foundation (IFF), told Al Jazeera that these particular ads are tagged with #aiart, so the content creators are somehow He said there is a need to recognize attempts to promote AI images. Restricting disclosures to comply with Meta's recently updated policies requiring disclosure of AI.
AI used to drive “strategic narratives”
As Al Jazeera previously reported, Indian election campaigns are slowly using AI-generated images and videos, including using AI to resurrect dead political leaders and using them as campaign art on party official accounts. It is a means of spreading the word. Despite the threat of AI-generated election misinformation, deepfakes aren't just being used to deceive voters. Instead, generative AI is employed to construct the narrative.
“While disinformation is undoubtedly a serious issue that requires our attention, it distracts from the more potential ways in which AI-generated content could help advance political parties’ strategic narratives.” It can also lead to problems,” Sharma from the University of Oxford told Al Jazeera after the AI study. image advertisement.
“It’s not just that AI is ‘tricking’ voters into believing things that are blatantly false. Rather, AI is enabling the creation of content that is more creative and can leverage more innovative cultural references. “This makes political propaganda more interesting, more shareable, and allows for wider circulation,” he added.
For example, the most popular AI image posted by the Hokage Modi Sama page is 'Our Saffron Superhero'. It shows Prime Minister Modi wearing a saffron-colored kurta pajama and a flowing cloak walking under the banner of Om, a Hindu spiritual symbol. The AI image has received nearly 2 million likes, with comments such as “Who wants India to become a Hindu nation?” In the comments section of Instagram. Most images recast the Indian prime minister as a “savior,” “saffron-clad guardian of India's future,” or “symbol of Hindu renaissance.”
Sharma pointed out that mythologizing political leaders or featuring them in Hindu epics is not new, but it is being done more subtly by AI.
“These AI-generated images of Narendra Modi appear to be the latest iteration of the Bharatiya Janata Party’s long-standing strategy. [Modi] As a wise and intelligent leader,” Sharma said.
Hokage Modi Sama's Instagram and Facebook pages do not have clear disclaimers regarding direct affiliation with Prime Minister Modi's Bharatiya Janata Party (BJP), but the pages contain pro-Hindutva and pro-BJP content. It has a history of publication. Previously, this page was image It dressed up Elon Musk, the owner of social media platform X, as a Hindu monk and captioned it, “Our fellow Bakht Elon,” to show that Musk is a Hindu nationalist.
Multiple attempts to contact the administrator of Hokage Modi Sama were unsuccessful. In the anime “NARUTO”, “Hokage” is an honorary title given to the leader of a village, and “sama” is a Japanese way of saying “teacher” with respect, so this page is roughly translated into “leader Modi”. means 'lord'. The Instagram handle, which has 130,000 followers, is currently selling products such as T-shirts, coffee mugs and diaries based on Saffron's superhero AI image.
Globally, AI imagery is being used in political campaigns. In March, supporters of US presidential candidate Donald Trump's campaign used a fake AI image of him posing with black voters in a bid to provide “strategic support” to the black community. In Argentina, two presidential candidates used AI-generated images to boost their popularity and attack their opponents. And in Indonesia, President-elect Prabowo Subianto, once a feared military dictator, used AI imagery to rebrand himself as a “cute grandpa” during his election campaign.
In India, ahead of the Telangana state elections in December 2023, fact-checking broadcaster BOOM Live announced that local political leader K. Chandrashekhar Rao will be launching a free meal scheme for students. It was reported that the AI images were incorrectly depicting the event, but Rao had never participated. .
Al Jazeera's December 2023 review of the meta ad library found that the Mana Telangana page posted multiple fake AI images of India's opposition leader Rahul Gandhi posing with children and farmers on Instagram. It was revealed that he had shared it on Gram and Facebook. Farmers make up India's influential voting bloc and have frequently protested against Prime Minister Modi's market-friendly farm laws.
The fake image of Gandhi with the farmers was accompanied by a caption that read, “Rahul Gandhi empathizes with the farmers and promises unwavering support and a promise to address their concerns individually.” The sponsored post included the hashtags #aigenerated and #aiimages to indicate it was generated by AI, and received an estimated 7,000 impressions.
West vs Global South
There are also disparities in how Western and Global South platforms approach the AI imaging trend ahead of elections.
For example, to reduce the potential for misuse of AI images ahead of the US presidential election in November, popular AI image generation company Midjourney announced in March the creation of fake images of President Joe Biden and President Donald Trump. was prohibited.
However, no such measures have been taken in India, raising “concerns about the fair enforcement of policies by technology platforms across different regions,” Babers Orsek, managing director of Logically Facts, told Al Jazeera. Told. Logally Facts is part of Meta's third-party fact-checking program and part of India's Anti-Misinformation Alliance.
Al Jazeera's attempt to generate an AI image of Modi and Gandhi shaking hands on Midjourney was successful, but a “forbidden prompt was detected” in the prompt that generated the image of Modi and Trump shaking hands. A notification was displayed.
“This decision highlights the disparities in how such moderation policies are applied globally, especially in the context of the Global South, including countries like India, which is facing the most important elections of the year. ” Olsek said.
Al Jazeera emailed Midjourney about the discrepancy in moderation, but has not yet received a response.
This is similar to how social media companies such as Facebook and Twitter are biased towards the West and often ignore perspectives from the Global South, as their policies for keeping their platforms secure added.