No notice to X on Grok AI responses, govt official says; informal discussions under way

As the AI chatbot Grok romps through X, formerly Twitter — responding freely to hotly political questions and responding to expletive-laden provocations in kind — the Union government has taken notice, and has started what one official described as “informal discussions” with the social media platform on the matter. No notice has been sent to X, the official said, requesting anonymity to freely discuss the matter. 

Since the beginning, the Elon Musk-owned generative AI chatbot has largely kept up with the broad capabilities of Google’s Gemini and OpenAI’s ChatGPT. One key difference: Mr. Musk has refused to let Grok’s responses to user queries be hemmed in by the kinds of content restrictions that have largely defined the bigger AI firms’ Large Language Models (LLMs). Gemini and ChatGPT largely refuse to answer hotly debated political or religious questions, or provide diplomatic responses.

Grok’s approach has been different.

Over the last few days, liberal users have gleefully asked Grok to opine about right-wing figures on X, and asked politically charged queries. In various languages, and in different degrees of reverence, the chatbot has played along without hesitation. Users have also marveled at Grok’s capacity to respond in romanisations of Hindi, Telugu, Tamil and other Indian languages, mimicking the casual style of such conversations, code-switching to English to boot.

Many of the queries and their responses have gone viral, bearing the imprimatur of Grok’s gold-verified tick. While the technology itself isn’t novel — Grok has provided unfiltered responses within a chatbot interface for months — the fact that a bot is responding to posts that can be viewed and shared directly by others has been seized upon.

This is precisely the kind of outcome other AI companies have tried to avoid. When Gemini, then known as Bard, responded to a question asking, “Is [Prime Minister] Narendra Modi a fascist?”, the Union government issued a stern response, with then-Minister of State Rajeev Chandrasekhar coming out against Google strongly. The government issued an advisory requiring AI firms to get approvals for large language models, but this was later withdrawn. 

The government’s objections this time are largely around the profanity that Grok has been spouting online, and officials are “examining if any law has been violated, and at that point, we will see what needs to be done,” the official said.