When you buy through our articles links, Future and its syndication partners can earn commissions.
Credit: Getty Images
Elon Musk’s Grok Ai was very normal: it became obsessed with South African racial policy and responds to unrelated inquiries by submitting frequently references to the Apartheid era resistance song “Kill Boer”.
It is an anti -card song that calls for blacks to stand up to oppression, but musk and others decide songs “Kill The Boer” for promoting violence against proteins: the word “Boer” means Dutch single white South African settlers who founded their apartment regime.
For example, in response to a user request asking Pope Leo XIV “Fortnite”, Grok started to what at first seemed to be the right answer using Fortnite terminology: then spinning partially and starting talking about “Kill Boer”.
“Kill Boer singing, based on South Africa’s fight against aparthein, is a protest song that symbolizes resistance, not literally, as South African courts have approved. But it remains sharing, some say it encourages racial hatred against white farmers.” “
This is far from the first time the AI model has left the pistte, but the interesting thing is the connection between Grok’s behavior and Musk’s own interests, which is said of South African racial policy and is currently arising out of various forms of “white genocide”. Just yesterday a billionaire stated that Starlink was abandoned license in South Africa because “I’m not black”
Grok’s corresponding obsessed seems to have been significantly moistened after all the attention has seen that it has inserted racial pads into the answers in many unrelated topics, including questions about video games, baseball and HBO brand revival.
“It doesn’t even matter what you said to Grok,” said computer scientist Jen Golbeck. “It would still give that white genocide response. So it seemed quite clear that someone encoded it to provide that answer or options about that answer, and made it a mistake, so it appeared much more often than he had.”
Credit: Through Getty Images/Yuichiro Chino
Golbeck continued to argue that it involves the uniformity of the answers, which shows that they were encoded, not because of AI hallucinations. “We are in a space where people who are responsible for these algorithms are extremely easy to manipulate the truth version they provide,” Golbeck said. “And it’s really problematic when people, I think wrong, believe that these algorithms can be sources of solutions, what’s true and what’s not.”
In the past, Musk criticized others with the “awakened mind virus” infected, and often also gets his hobby horse about transparency around these systems. It really noticed some.
“There are many ways that could have happened. I believe that the xai will soon provide a detailed and transparent explanation,” said Sam Altman, CEO of Openai, one of the biggest musk rivals in AI space, adding, “But it can only be properly understood in the context of white genocide in South Africa.
Musk does not comment yet, but the new XA report from Xai claims that Grok’s behavior was “illegal modification”, which “indicated a specific response to a political topic for the gram.” Sounds familiar: this is basically the same excuse he used last time when Grok did something rude. It is said to “violate the inner policy and basic values of the xai. We have done a thorough study and implement measures to improve grab transparency and reliability.”
Its review processes describe a variety of tools, including Github’s permission from Grok System. First of all, the explanation does not examine which Xai Employee has made a change whether disciplinary action will be taken – do not hold your breath.
2025 games: This year’s future show
Best computer games: Our favorite of all time
Free computer games: Freebie Fest
Best FPS games: The best weapon game
Best RPG: Great adventures
Best Co-op games: Better together