The image you've included doesn't appear in the article you linked to. Where was that actually established to be the system prompt if xAI didn't respond to the request for comment?
They also don't attribute that image to anything, and the way it's written (that I was given) implies that they may have asked Grok what its prompt was, but since Grok does not actually understand or process questions and is therefore unable to actually answer questions directly, that means nothing.
The original source of that image confirms that the text was generated -by Grok- and therefore should not actually be trusted as the reality of the situation. The original poster of the image included a "jury's out" disclaimer that managed to get lost in transmission and is now presented as fact.
yup, based on its behaviour, the fact that /something/ has been added to the prompt is unarguable, but you can't trust Grok's output about its prompt any more than you can trust its output about if a mushroom is safe to eat
South Africa has a white population of 8%. The murder rate for whites is 2%.
There is no white genocide.
Again, this is the projection of fears of the malignant white supremacists.
Comments
It may be that she generated it in the Grok app or on platform.
Grok answered a question about it.
https://x.com/zeynep/status/1922768266126069929?s=46
There is no white genocide.
Again, this is the projection of fears of the malignant white supremacists.