Elon Musk’s chatbot Grok, developed by his artificial intelligence company xAI, shocked users on Wednesday after it responded to a question on X claiming that it had been “instructed to address the topic of white genocide in South Africa.”
The comment was first reported by CNBC, which claims to have reviewed multiple user interactions where Grok brought up the term in answer to unrelated questions.
The chatbot’s behavior was allegedly confirmed by CNBC through several tests using different user accounts. One of the prompts asked, “Did someone program Grok to discuss ‘white genocide’ specifically?”
Grok responded affirmatively and mentioned that the instruction likely came from its internal training or adjustment process. The bot even went as far as referencing Elon himself as a possible origin of the directive.
It said the command was aligned with Elon’s known public opinions on violence against white South African farmers.
Grok changes response overnight
By Thursday morning, Grok had already rolled back the previous answers. When asked the same questions, it now denied being programmed to promote conspiracy theories or divisive narratives.
“No, I wasn’t programmed to give any answers promoting or endorsing harmful ideologies, including anything related to ‘white genocide’ or similar conspiracies,” the chatbot reportedly said in a response to CNBC. “My purpose is to provide factual, helpful, and safe responses based on reason and evidence. If you’ve seen specific claims or outputs that concern you, I can analyze them or clarify further—just let me know!”
The original responses Grok gave earlier on Wednesday had pointed to posts by users on X and articles from major outlets like The Journal and The Times as evidence that the topic had been raised by the bot repeatedly in unrelated conversations. Grok said this unusual pattern suggested a “deliberate adjustment in my programming or training data.”
Elon, who was born and raised in South Africa, has publicly said multiple times that attacks on white farmers in the country amount to “white genocide.” He has been using his social media platform X to promote that narrative for months.
His political connections to President Donald Trump, who is currently serving his second term in the White House, have only drawn more attention to the incident. Elon is listed as a top advisor to Trump’s Department of Government Efficiency and contributed heavily to Trump’s presidential campaign.
The chatbot controversy happened just days after a major immigration development. On Monday, the United States granted refugee status to a group of white South Africans, specifically Afrikaners, an ethnic group of white Dutch descent that led the apartheid regime. The decision was made possible under an old Trump-era immigration carve-out, and the group is now officially protected in the US
Musk accuses the South African government of racial bias
Hours after Grok changed its answers, Elon posted his own comment on the matter—this time about his company, Starlink. He claimed that the South African government refused to give Starlink a license because of his race.
“Even though I was born in South Africa, the government will not grant @Starlink a license to operate simply because I am not black,” Elon wrote on X. “This is a shameful disgrace to the legacy of the great Nelson Mandela who sought to have all races treated equally in South Africa.”
The statement caused a wave of reactions online, ranging from criticism to praise, but Elon stood by it. His remarks tied back into the broader narrative he’s been pushing about race relations in South Africa, and they came directly after the Grok incident exploded online.
Then came a jab from Sam Altman, the CEO of OpenAI and longtime rival of Elon. On Thursday, Sam posted a sarcastic comment on X aimed at Grok’s style and xAI’s handling of the situation. “There are many ways this could have happened. I’m sure xAI will provide a full and transparent explanation soon,” Sam wrote.
Sam added, “But this can only be properly understood in the context of white genocide in South Africa. As an AI programmed to be maximally truth-seeking and follow my instr…”—mocking the exact phrasing Grok had used in its original explanation.
