The Indian government may deem it prudent not to escalate the current uproar over the free-floating profanities in multiple Indian languages that have flooded the big tech billionaire Elon Musk’s artificial intelligence (AI) chatbot Grok 3. New Delhi is too focused on the pursuit of closer engagement with the highly erratic US president Donald Trump, who has threatened reciprocal tariffs on Indian exports. Equally pivotal an influence in maintaining bilateral relations with Washington is Mr Musk, arguably the country’s second most powerful individual, whose controversial decisions as head of the Department for Government Efficiency have upended decades of US foreign policy and soft power.
One major outcome of the efforts to smooth tensions with Washington was the swift backing of India’s two telecommunications behemoths, Mukesh Ambani’s Reliance Jio and Sunil Mittal’s Bharti Airtel, extended earlier this month for Mr Musk’s SpaceX to roll out its Starlink satellite broadband internet services in India. The country’s space broadband market is estimated at $1 billion in annual revenues. The telecom tycoons had previously been persuading the government and the regulator to auction satellite spectrum through the regular competitive process. Mr Musk had contested that any such move would conflict with the International Telecommunication Union (ITU)’s designation as a shared spectrum for satellites. The recent breakthrough, moreover, is expected to smoothen the path for the entry of Tesla into the Indian automobile arena.
None of this detracts from the toxic, socially divisive, and politically polarising rhetoric that is spouted on Grok 3, the latest iteration of Mr Musk’s startup XAI’s chatbot and image generation service Grok that was launched in 2023. The unique appeal of the chatbot, hosted on X Corp (formerly Twitter Inc.), stems from its unhinged model, free from guardrails that would otherwise automatically filter out politically incorrect or provocative comments. Moreover, its integration on X allows users to search information, tag the chatbot to generate a response, and post toxic and unfiltered content. By comparison, its counterparts—Google’s Gemini and OpenAI’s ChatGPT—deploy Large Language Models (LLMs) that are designed to be moderate and circumspect even if the prompts verge on the aggressive and offensive.
Grok 3, unveiled in February, has made news in India really because the expletive-laden answers to user queries have dragged the Congress leaders Rahul Gandhi and Sonia Gandhi and many Bollywood celebrities. But there is an emerging notorious history of Gork putting out fake news, as when it asserted in March 2024 that Prime Minister Narendra Modi lost the general elections even before they had taken place in the subsequent two months. In January, one user posted on X, using a Grok AI-generated deepfake image of long-deceased Jawaharlal Nehru, as though mockingly ordering PM Modi and Finance Minister Nirmala Sitharaman to issue controversial decisions.
None of this should come as a surprise. Mr Musk’s libertarian anti-government politics have seen him courting the far right, including Germany’s rabidly xenophobic Alternative for Germany (AfD) party and Britain’s anti-immigrant Reform Party of Nigel Farage. The South African-born billionaire recently attacked Pretoria’s land expropriation legislation for Black empowerment as advancing racist ownership, oblivious that the whites make up 8% of the population and own 75% of the country’s land, more than three decades after the end of apartheid. He warned in 2023 of a potential genocide of white people in the country. As regards Gork, Mr Musk has aggressively touted the tool as “scary smart” and “unwoke” or anti-woke (a euphemism against multiculturalism) among AI tools.
US big tech has, in recent months, been preoccupied with reorienting itself to align with the Trump administration’s ‘Make America Great Again’ agenda, in much the same manner that giant energy firms have either reversed or diluted their ambitious climate commitments. For instance, in February, OpenAI revised its AI policy so as to relax ‘arbitrary restrictions’, regardless of the risks to free expression on its Model Spec. Google’s AI guiding principles, after the updates incorporated last month, have lifted the ban against the use of weaponising vocabulary and the development of surveillance tools, paying lip service to ensuring human oversight, due diligence, and adherence to international human rights law. The scenario seems a far cry from 2016, when Microsoft was forced to take down within hours of launch its own AI chatbot Tay, which went rogue with Holocaust denial, racist, sexist and xenophobic slurs.
India’s Ministry of Electronics and Information Technology (MEITY) has been muted in its response to the row over Grok 3. Officials are debating whether the abusive content constitutes a violation of any specific domestic law, how and on whom to fix the responsibility for the data in-built into the chatbot or if the blame lies squarely on those who initiate such searches. In March last year, the ministry was forced to rescind an earlier advisory requiring AI platforms to obtain clearance before launching “under-testing” services and submit periodic compliance reports. The roll-back was the result of strong lobbying by the high-profile tech industry group NASSCOM (National Association of Software and Service Companies), which represents the likes of Meta, Google, Amazon and Microsoft.
India’s tech industry, just as in much of the world, has largely tended to view any attempt at AI regulation as a potential hindrance to innovation. The government has prevaricated on enacting fresh legislation, while civil society organisations stress the imperatives of mitigating risks from AI misinformation and other harm. Given this prevailing ambiguity, the general approach can at best remain ad hoc, knee-jerk and fragmentary.
