The original vision of Grok was for it to be the less politically correct alternative to all the other LLM AI chatbots out there. But perhaps xAI, or Elon Musk himself, though that it didn’t go far enough, so over the weekend it was given additional prompts by its makers to lean more into its politically incorrect side. And so it did, maybe a little too much, as the company has had to not only clean up the mess the chatbot left behind, but also remove some prompts to be more politically incorrect.
Musk announced on Friday that Grok got improved “significantly”, though didn’t mention details beyond the claim. Then xAI got to work adding the aforementioned system prompts, which are publicly posted on GitHub. More recently, the chatbot has been posting some pretty extreme statements, to the point that xAI had to step in to not only remove said posts, but also prevent the chatbot from making more text-based posts.
Of course, the internet never forgets, so some of these posts got screengrabbed before they were deleted, and these screenshots are now making their rounds. xAI has said that it is “aware of recent posts made by Grok and are actively working to remove the inappropriate posts”. The company also claims that it “has taken action to ban hate speech before Grok posts on X”.
If you’re curious, you can check the GitHub links below to see what prompts were added to Grok, and what ended up being removed as part of the aftermath. Not too surprising, since the prompt was taken to its logical extreme, and then some.
(Source: @MikeIsaac / X, GitHub [1], [2])
