X’s Grok AI revives concern about deepfakes ahead of Delhi Assembly election

AI deepfakes in Indian politics spark debate over ethics and impact, raising concerns about misinformation and satire.

The Aam Aadmi Party posted an Artificial Intelligence-generated deepfake of B.R. Ambedkar supposedly blessing former Delhi Chief Minister Arvind Kejriwal last month, prompting an AI-generated response from the Bharatiya Janata Party in return.

In the run-up to the Delhi Assembly election, the exchange has reignited the debate over the use of AI in political campaigning, and the role it plays in elections. This use of deepfakes has been growing since Grok — the AI chatbot and image generation service offered by X, formerly Twitter — became available to the general public. Unlike the policy followed by other AI chatbots, X’s owner Elon Musk has decided against prohibiting imagery based on real life political figures, leading to a mushrooming of such content on Grok.

For instance, an X account riffs on the constant blaming of India’s first Prime Minister Jawaharlal Nehru for modern ills by generating images of the long-deceased Nehru ordering current PM Narendra Modi and Finance Minister Nirmala Sitharaman to take unpopular decisions. The account, @The_Nehru, has gained over 20,000 followers. The account is “based on the recent parliamentary debates where Modi started blaming Nehru, which he does most of the times,” its creator, going by the alias JLN, told The Hindu over direct messages on X. “So I got the idea to create [a] parody account and mock the statements.” 

Adhiraj Singh, a comedian who writes on Indian humour, and one of the co-contributors to a satirical page called Humans of Delhi (aping Humans of Bombay and other such pages), was skeptical about AI being a sustainably funny mainstay for a crop of accounts. “Satirical pages pretending to be politicians aren’t new, but AI tools do make it easier for them to flood our timelines with trash that ultimately make any satire or commentary meaningless and inseparable from any other kind of noise,” Mr. Singh said. “I feel it really depends on who is using it for what.”

He added that there were concerns with this kind of content becoming more common: “Satire, misinformation, and hate speech being used interchangeably with no accountability. People and even news sources mistaking ‘satire’ pages for genuine news isn’t even news any more. It’s not sustainable, but here we are, in the post-singularity AI slop pit.”

Indeed, the use of deepfakes in this way is not restricted to political parties and parody accounts alone. One political commentator posted a picture of Congress leader Rahul Gandhi resembling the businessman and entrepreneur George Soros, with a caption calling him Mr. Gandhi’s “mentor”. 

In a report on generative AI deepfakes and India, the disinformation-focused startup Logically wrote that concentrating “on specific kinds of content alone in assessing whether there will be any impact can obscure the way that disinformation campaigns operate. The consequences actually lie in the cumulative effect of the content appearing endlessly in a variety of different fora.”