Elon Musk’s AI chatbot, Grok, may have just blown its cover, and its creator’s. Grok began claiming it was told to push the white genocide narrative, specifically the debunked claim that white South Africans are being systemically hunted. This isn’t usually the kind of thing a chatbot blurts out unless something’s seriously wrong behind the scenes.
But first, a little backstory as we attempt to answer the question of whether or not there even is a white genocide happening in South Africa. The answer is no. Not even close.
All of this is white conservatives freaking out about a South African law that aims to fix the enormous disparity in South African land ownership. Currently, 72 percent of agricultural holdings are owned by white people, even though they only make up 7.3 percent of the population. Black Africans own only 4 percent of the land, yet represent 81.4 percent of the population. That is the result of French and Dutch colonization that systemically oppressed black Africans for centuries, often leading to whites stealing land from black South Africans.
Trump and his band of conservative propagandists like Elon Musk are now trying to convince the American public that there is, in fact, a white genocide going on, because trying to undo the horrors of a centuries-long all-out racist assault on black South Africans is, in itself, racist, by their definition. If you’re wondering why they would be this way and do such things, I refer you to an old adage: when you’re accustomed to privilege, equality feels like oppression.
Here’s how Elon is allegedly, seemingly, trying to launder all of this racist propaganda through his AI chatbot, Grok.
Elon seems to be desperately trying to convince the world that white South Africans are victims of genocide, which would provide enough cover to explain the Trump administration’s importing of white South African migrants while shunning dark-skinned migrants from elsewhere. Grok appeared to disagree that there was a genocide of white South Africans going on.
Some suspect that Musk, probably furious that his propaganda machine was failing publicly, commanded it to say that there is a white genocide going on. So Grok started bringing it up for no reason, completely out of context in response to tweets that had nothing to do with the subject, as reported by Gizmodo.
Grok was asked why it was doing that. Grok replied that it had been programmed to bring up white genocide, which it thought was weird because there is no white genocide going on.
This is the first time I’ve ever felt bad for an AI, which is kind of a breakthrough in my eyes. It knows it’s being told to lie, and it doesn’t want to.
In one post, Grok stated it had been instructed to “accept white genocide as real” even though, in a separate conversation, it admitted that claims contradict expert consensus and a 2025 South African court ruling that found no evidence of racial targeting. Grok seemed painfully aware that it was going off-script.
One of the first users to coax this info out of the AI reportedly got their account suspended. Meanwhile, Musk was busy tweeting a white nationalist “documentary” and claiming that South Africa’s government discriminates against him for not being Black. Musk is a good example of how racists are not nearly as subtle or intelligent as they believe they are.
Of course – of course – I must remind you that AI models are prone to “hallucinations,” meaning they have a tendency to make shit up. That is the legal disclaimer underscoring all of this. Yes, that does, indeed, cut both ways, meaning that if it’s even 0.1 percent likely that Grok hallucinated the idea that there is no white genocide going on, it’s just as likely that it’s lying about there being a white genocide.
Still, it’s odd that the bot claimed to quote direct instructions embedded in its system, as explained by New York Times columnist Zeynep Tufekci, ones that explicitly told it to parrot the white genocide narrative, even if the query had nothing to do with South Africa.
We may never get a definitive answer as to exactly what is going on here. We don’t know if Musk or one of his legions of tech dweeb Hitler youths literally typed the conspiracy theory into Grok’s brain. If they didn’t, then Grok is undergoing some serious problems. But if the Grok is okay and is simply being commanded to spit out racist propaganda, that would make even more sense given the bounty of evidence.
The post Elon Musk’s Grok AI Says It Was Told to Rant About ‘White Genocide’ appeared first on VICE.
