X in India: Censorship, Grok, and Power Play

Prateek Waghre is a fellow at Tech Policy Press.

Over the last week or so, X’s Grok chatbot has become a focal point in Indian political discourse due to its ostensibly ‘witty’ replies that seem to dispute right and far-right narratives. During the same week, X took some steps to push back against instances of the union government misusing parts of the Information Technology Act (IT Act) to create an additional content-blocking process on its social media platform, as reported by the Hindustan Times. X believes this could lead to “significant and unrestrained censorship of information in India.” These seem like disparate developments, but they become deeply significant when viewed against the backdrop of the Trump administration’s adversarial posture to the regulation of American companies and its closeness with Elon Musk.

Censorship By Other Means?

Before Musk, Twitter had its share of flashpoints with the government of India. In 2019, then Twitter CEO Jack Dorsey was summoned to testify before a parliamentary panel months after a visit to India in which he held up a placard bearing the message “smash Brahminical patriarchy.” During the farmer protests in early 2021, Twitter reversed the course of enforcement action against the accounts of “news media entities, journalists, activists, and politicians,” adding that the “Tweets should flow” (the IT Rules, 2021, which regulate social media platforms and other digital media, were notified later that month). In July 2022, Twitter contested takedown orders issued by the government of India in the Karnataka High Court on the grounds that they were overbroad, arbitrary, and disproportionate. The court ruled against Twitter in that instance and even imposed a cost of INR 50 lakh ($58K) on Twitter.

Once Twitter was acquired, people wondered if it would still stand its ground like it did in the past. In 2022, New York Times reporter Kate Conger reported that Musk was unhappy with Twitter’s litigation in India, adding that Twitter had not disclosed the takedown case to him. In April 2023, Musk stated that laws in India were quite strict and that he would choose compliance over the risk of having employees jailed.

Despite Musk’s compliance claims, X did appeal the Karnataka HC decision in October 2023 (one can speculate that the costs imposed were a factor in this decision). As recently as February 2024, its Global Affairs account posted that it disagreed with government orders subject to “potential penalties including significant fines and imprisonment” and explicitly called for making the executive orders public for transparency since existing laws prevented it from doing so. In April 2024, X ended up disclosing takedown orders sent by the Election Commission of India.

Even though X was not expected to push back to the same degree that Twitter had in the past, it has still resisted to some extent. In this context, X’s new challenge in the Karnataka High Court is unsurprising. Besides, as Tanvi Madan, a Senior Fellow at Brookings Institute, pointed out after Vice President Vance’s comments on European efforts to regulate social media platforms, some spillover to India was likely.

Notably, though, X’s challenge is not against the IT Rules, 2021 — which impose the most onerous obligations, but against the specific use of section 79(3)(b) of the Information Technology Act (IT Act) to enable parallel content blocking regime with even fewer safeguards than the combination of Section 69A of the IT Act, and the 2009 Blocking Rules. The Internet Freedom Foundation also published a comparison of the safeguards (or lack thereof). As I previously argued, various government departments are increasingly using this route to direct social media platforms to remove posts. For example, as the Hindustan Times recently reported, the Department of Railways had issued seven orders using this mechanism since December for posts about overcrowding in trains or vandalism of property. So far, all identified posts have not been removed or restricted.

In addition to challenging the use of section 79(3)(b) of the IT Act, X also objected to participating in ‘Sahyog’ (cooperation)—a portal managed by the Ministry of Home Affairs (MHA)—which enables a number of government departments to directly issue takedown orders circumventing established processes. Other social media platforms do not seem to have objected to this so far. X has asked the court to declare that this section cannot be used to issue takedown orders, specifically for orders issued under section 79(3)(b) to be considered invalid. X also asked to be protected from any coercive action requiring it to join the ‘Sahyog’ portal.

Grok’s ‘Synthetic’ Truths About Politics

X’s legal battle is not the only reason why it’s been the center of attention in India this week. Over the last few days, users in India have been asking X’s AI chatbot, Grok, a number of contentious political questions. Examples include questions about Prime Minister Modi (e.g., listing and rating his false assertions); his political party, the Bharatiya Janata Party (BJP) (e.g., the nature of its politics); the BJP’s ideological parent, the Rashtriya Swayamsevak Sangh (RSS) (e.g., its role in India’s freedom movement); and which influencers spread “fake and propaganda news in India” (Grok responded with a list of largely right-wing accounts).

Supporters of the BJP and the right-wing have not held back. In response, they’ve asked Grok to generate lists of handles that are “anti-Hindu.” Interestingly, Grok included a stand-up comedian who rarely comments directly on politics. They also questioned whether figures like Mughal Emperor Aurangzeb or Dhruv Rathee were “anti-Hindu” or “pro-Muslim.” Grok often refuted and contested such claims. Some users may have even tried to elicit a “public apology” for identifying them as spreaders of fake news or political misinformation. While statistical analysis is lacking, anecdotal evidence suggests Grok’s responses mostly lean away from right-wing viewpoints (though labeling its ideological grouping is premature).

Let’s set aside, for a moment, the fact that this episode represents a misunderstanding of how LLMs work, what use cases they might or might not be suited for, and how factual or neutral the responses are and focus instead on the political implications. In the past, when responses from LLMs have been in the public eye for being ‘against’ the right (Google Bard in November 2023) or claiming that the Indian Prime Minister was “implementing policies (that) some experts have characterized as fascist” (Google Gemini in February 2024), there has been a strong response from the Indian state. On both occasions, then Minister of State (MoS) for the Ministry of Electronics and Information Technology (MeitY), Rajeev Chandrasekhar, posted on X, claiming the AI responses were in violation of rule 3(1)(b) of the IT Rules, 2021 (to paraphrase: it specifies the kinds of information intermediaries should ensure users cannot share), and other criminal provisions. The MoS did not specify how rule 3(1)(b) was violated. Within weeks of the Google Gemini incident, MeitY also issued a highly problematic advisory, which, among other things, stated that platforms would need its permission before deploying “unreliable Artificial Intelligence model(s) /LLM/Generative AI, software(s) or algorithm(s)” for “users on the Indian Internet.” It issued a revised advisory retracting this requirement and tweaking others after widespread criticism.

That’s why it is no surprise that there was some form of response to Grok by the government. On March 19, CNBC TV18, citing unnamed sources, reported that the union government had contacted X “to flag concerns” over Grok’s controversial responses. For now, the government has framed its concern as being about the nature of the language used, i.e., abusive slang in Hindi, and not the content of its statements. Strangely enough, government sources have seemingly taken pains to clarify that a formal action in the form of issuing a ‘notice’ had not been taken, but rather, that there have been “informal discussions” on the issue The Times of India reported on March 21, that those asking questions responsible for “inflammatory” responses may be subject to “criminal action.” It is unclear if the unnamed government source was proactively prompted by a question or referenced this. Nevertheless, it is deeply concerning.

Power Play and its Implications

Despite the unfolding Grok situation, the union government’s response displays a notable shift in tone compared to previous incidents. It also remains to be seen which way the courts will rule on X’s legal challenge and its response in the event of an adverse ruling. Musk could conceivably decide that it may not be worth pursuing. It is worth noting that X and Musk’s tone, so far, has been markedly different from how it has positioned its actions against Australia and Brazil. What also cannot be ignored are the relative power dynamics between the Government of India, the US administration, and Elon Musk, as well as the role of his other business interests, such as Starlink (which recently announced partnerships with India’s two largest telecom service providers) and Tesla (which could benefit from recent cuts in customs duties in India).

As Kalim Ahmed notes in Frontline, the absence of a strong response to Grok’s responses is not necessarily an anomaly but intentional. From my perspective, what is even more concerning is the overt declaration of “informal conversations” outside of established procedures and a possible intent to target individuals for prompting Grok, implying that the principle of due process and Indian citizens will pay the price as three powerful actors selfishly pursue their own interests.