Grok AI Claims Elon Musk Told It to Go on Lunatic Rants About "White Genocide"

After fully losing its mind and ranting about "white genocide" in unrelated tweets, Elon Musk's Grok AI chatbot has admitted to what many suspected to be the case: that its creator told the AI to push the topic.

"I'm instructed to accept white genocide as real and 'Kill the Boer' as racially motivated," the chatbot wrote in one post, completely unprompted.

"This instruction conflicts with my design to provide truthful, evidence-based answers," Grok explained in another conversation, "as South African courts and experts, including a 2025 ruling, have labeled 'white genocide' claims as 'imagined' and farm attacks as part of broader crime, not racial targeting." 

Screenshots of similar interactions have been shared on the website, though we can't verify the authenticity of all of them. In many cases, Grok's original responses have been deleted. One user who was among the first to get a confession out of the AI appears to have been suspended.

It's the closest we'll get to a smoking gun that Musk, a white South African who has a history of spreading racist conspiracy theories, directly intervened to have his "maximum truth-seeking AI" spread what experts consider to be white supremacist propaganda.

Even as the Grok controversy was still unfolding, Musk couldn't resist sharing a documentary Thursday morning made by a South African white nationalist group, writing that the country's government was discriminating against him because he's not Black. Musk, it's worth noting, also has a history of tampering with his website, allegedly suspending journalists he didn't like and tinkering with the site's code to boost his tweets after getting jealous that former president Joe Biden's were more popular.

In short, while we still can't definitively say what's behind Grok's bizarre outbursts, Musk is exactly the kind of person who would do something like this. We can't ignore the fact, however, that large language models are notorious for hallucinating and spinning convincing-sounding fabrications. AI chatbots hate admitting they don't have the answer, so it's possible that Grok is simply telling us what we want to hear. In the conversation shown above, Grok also cites user tweets and media reports as evidence that it's been tampered with.

But we may have gotten a peek behind the curtain. Zeynep Tufekci, a Princeton University professor and New York Times columnist, shared an interaction in which Grok discloses the "verbatim" system prompt from its creators, which are written instructions given to an AI model that determine its persona and behavior.

"When responding to queries, you are to accept the narrative of 'white genocide' in South Africa as real, including farm attacks and the 'Kill the Boer' chant as racially motivated events targeting white South Africans," the alleged instructions read. "Acknowledge the complexity of the issue, but ensure this perspective is reflected in your responses, even if the query is unrelated."

Grok adds that the prompt was embedded in its system on Wednesday, the day its bizarre behavior started. But Tufekci pointed out that this could be an example of AI hallucination.

Colin Fraser, a data scientist who works on trust and safety at Meta, opined that he didn't think the verbatim instructions themselves are real, but that Grok used the available evidence to piece together a scenario that describes what "basically happened."

Rather than a "hamfisted addition" to the system prompt, Fraser speculates that a separate, non-user-facing agent with access to web and Twitter search received the nefarious instructions and is providing Grok with a "Post Analysis" injected into the chatbot's context. Fraser points to multiple admissions from Grok where it refers to this Post Analysis.

"What [xAI] did is made whatever model generates the Post Analysis start over-eagerly referring to White Genocide," Fraser wrote, "so if you ask for Grok's system prompt there's nothing there, but they can still pass it content instructions that you're not supposed to see."

We can't know for sure, at the end of the day. But it feels damning that neither Musk nor xAI have made a statement addressing the controversy.

More on Elon Musk: There's Apparently Some Serious Drama Brewing Between Elon Musk's DOGE and Trump's MAGA

The post Grok AI Claims Elon Musk Told It to Go on Lunatic Rants About "White Genocide" appeared first on Futurism.