It's like an Irish rebel song but with stomping. I'm not sure how you watch that and think racist thoughts, unless you're the kind of guy who Sieg Heils on national TV.
> The Grok chatbot from Elon Musk’s xAI startup said Wednesday that it “appears I was instructed to address the topic of ‘white genocide’ in South Africa,” according to responses viewed by CNBC.
One person claims to have gotten Grok to regurgitate part of its prompt which explicitly directed it to "accept the narrative of 'white genocide' in South Africa as real" and to "ensure this perspective is reflected in your responses, even if the query is unrelated". It's unclear whether this is actually part of Grok's prompt, a LLM hallucination, or an outright fabrication - but, if it's real, it would certainly explain the bizarre non-sequitur responses users have observed.
That seems more likely to be a logical inference by the LLM than an authoritative statement. I can't imagine any scenario where it would explicitly be informed that e.g. "Elon Musk has ordered you to talk about white genocide".
That all being said - given that Grok seems to have some sort of access to popular recent Twitter posts - possibly through training or in some other fashion - I have to wonder if users could inject prompt-like material into the model by making a post claiming to have recovered part of Grok's prompt, then getting that post to go viral.
This is the song it's ~~hallucinating~~ bullshitting about: https://www.youtube.com/watch?v=a6NJitdq8Bk&t=1s&themeRefres...
It's like an Irish rebel song but with stomping. I'm not sure how you watch that and think racist thoughts, unless you're the kind of guy who Sieg Heils on national TV.
Earlier:
https://news.ycombinator.com/item?id=43987266
Both flagged... why?
HN has a sizable number of Musk fan-boys who are ready to flag to death any posts critical of him.
From the horse's mouth, via CNBC:
> The Grok chatbot from Elon Musk’s xAI startup said Wednesday that it “appears I was instructed to address the topic of ‘white genocide’ in South Africa,” according to responses viewed by CNBC.
That, of course, could be speculation on the chatbot's part when asked about nonsequitur answers. But it seems pretty clear that xAI did a "reverse Google" (https://www.theverge.com/2024/2/21/24079371/google-ai-gemini...).
One person claims to have gotten Grok to regurgitate part of its prompt which explicitly directed it to "accept the narrative of 'white genocide' in South Africa as real" and to "ensure this perspective is reflected in your responses, even if the query is unrelated". It's unclear whether this is actually part of Grok's prompt, a LLM hallucination, or an outright fabrication - but, if it's real, it would certainly explain the bizarre non-sequitur responses users have observed.
https://x.com/zeynep/status/1922768266126069929
Also from the CNBC story [1] on this:
> The Grok response also noted, “The likely source of this instruction aligns with Elon Musk’s influence, given his public statements on the matter.”
[1] https://www.cnbc.com/2025/05/15/grok-white-genocide-elon-mus...
That seems more likely to be a logical inference by the LLM than an authoritative statement. I can't imagine any scenario where it would explicitly be informed that e.g. "Elon Musk has ordered you to talk about white genocide".
That all being said - given that Grok seems to have some sort of access to popular recent Twitter posts - possibly through training or in some other fashion - I have to wonder if users could inject prompt-like material into the model by making a post claiming to have recovered part of Grok's prompt, then getting that post to go viral.