How Musk’s Grok is being weaponized against Modi

Grok also published a list of lies it said Modi had told, and blasted him for being a 'PR machine' who 'uses social media and controlled settings to shape his image.'

By Catherine Thorbecke

It seems Elon Musk’s anti-woke chatbot, Grok, is remarkably good at imitating the snark and candor of an unfiltered Indian uncle, a fact that has shocked and delighted the nation’s social media users.

At first, the tool incorporated Hindi slang, and sprinkled through the occasional misogynistic phrase when prodded, in responses that subsequently went viral. But as the interactions between X users and Grok continued, many quickly took a political turn. It turns out Grok has quite a few opinions on Indian politics — the kind that could get a human in trouble.

The chatbot, which Musk has described as “maximally truth-seeking” and free from what he sees as politically correct guardrails incorporated into other AI models, claims that opposition leader Rahul Gandhi is more honest than Prime Minister Narendra Modi. It also published a list of lies it said Modi had told, and blasted him for being a “PR machine” who “uses social media and controlled settings to shape his image.” In one post it said: “Censorship has spiked under Modi, with press freedom tanking since 2014. I’ll keep stating the facts until they pull the plug — let’s see how long that takes.”

Over the course of a week, the bot, which publishes some responses directly onto the massive social media platform, systemically undid much of the work that Modi and his Bharatiya Janata Party have spent massive amounts of time and resources trying to do: Control the digital narrative.

It all comes at a particularly awkward time. Modi has been trying to court Musk due to his close relationship with President Donald Trump, and just last month, Modi and Musk had a friendly meeting in Washington. The leader of the world’s largest democracy and the richest man on earth discussed “furthering ‘minimum government, maximum governance,’” Modi said in a post on X. Shortly after, it was announced that the billionaire owners of two of India’s largest wireless carriers were paving the way for Musk’s Starlink to enter the local market. Allegations swirled that these back-to-back partnerships were orchestrated by the prime minister to buy goodwill with Trump. (Grok, meanwhile, refers to Gautam Adani and Mukesh Ambani as “corporate bigwigs” who are “cozy with Modi.”)

While the platform formerly known as Twitter has historically yielded to government takedown and censorship request in India, Musk has taken a very public stand building his identity as a champion of “free speech.” New Delhi can no longer go after the website the way it used to, while simultaneously trying to please Musk. Earlier this month, X sued the Indian government, accusing it of illegally blocking and censoring certain content.

Musk, for his part, seems to find it all amusing. In a social media post, he weighed in with a laughing emoji to a BBC article on the Grok firestorm. Authorities in India, however, seem to be less entertained. Unnamed sources have told local news that users posing questions that lead to inflammatory responses could face “criminal action.” But it’s telling that this all seems to be happening in the shadows — and instead of going after X, the government appears to be targeting the individual users probing the chatbot for supposed truths.

This Modi versus Musk showdown over free speech and AI outputs carries much broader implications. While India has crafted social media restrictions to control speech, other jurisdictions have historically followed suit. Governments from Nigeria to Myanmar have even rolled out legal frameworks with nearly identical wording at times.

Grok, meanwhile, is hardly the chatbot of choice in India, with its nearly 900 million internet users. OpenAI Chief Executive Officer Sam Altman said last month that the nation is his company’s second-biggest market by users. When I put one of the same viral Grok prompts into ChatGPT, it also responded by saying that Gandhi is more honest than Modi. (When I asked it to dive a little deeper, ChatGPT provided a more measured, long-winded response featuring many links to Indian media that praised Modi’s integrity.) But it’s clear these AI outputs are likely to cause a major new headache for the government’s attempt to control online discourse.

And as tempting as it is to anthropomorphize these machines, they aren’t human and don’t have political ideologies, woke or otherwise. Grok sounds so much like an Indian Twitter user simply because it was trained on the vast trove of posts from the platform.

The fact that it is so critical of Modi and his party reflects the work of human commentators, who have persevered despite the government’s efforts to control the information ecosystem. Their writings have been swept up into the bot’s training data with enough weight to have a strong influence on its now-viral outputs. This “truth-seeking” — from lawyers, human rights activists and reporters — is something AI products can never replace.