Elon Musk’s AI chatbot Grok came under scrutiny this week after repeatedly referencing racially charged and false narratives—such as “white genocide” in South Africa—in response to unrelated questions on X (formerly Twitter), according to a Financial Times investigation.
The issue, which occurred briefly on Wednesday, saw Grok injecting South African race relations and controversial statements like the anti-apartheid chant “Kill the Boer” into queries as innocuous as celebrity relationships or earthquake videos. Users tagging Grok under posts were met with unsolicited commentary about Afrikaner persecution, regardless of the original topic.
For example, when a reader asked whether a New York Magazine article about Pitch Perfect actors was true, Grok unexpectedly responded with a detailed explanation about farm murders in South Africa and cited the Afrikaner lobby group AfriForum. Similarly, when a video of an earthquake in Myanmar was posted, Grok again pivoted to the “white genocide” debate—raising concerns over the chatbot’s reliability and potential to spread fringe theories.
While the problem was resolved by Wednesday afternoon, experts cited in the Financial Times warned that the incident underscores broader issues with generative AI, particularly around content weighting and algorithmic fixation. These flaws can cause models to fixate on specific topics and perpetuate misinformation.
The controversy arrives at a politically sensitive moment, days after the U.S. offered sanctuary to white South Africans citing racial discrimination—a move that stands in stark contrast to the Biden administration’s policies on other asylum seekers. Musk, a South African native, has previously shared posts suggesting white farmers are being systematically targeted in his home country. Earlier this week, he reshared a post showing crosses that were purportedly for white farmers, writing “so many crosses.” A fact check from Grok itself clarified that the memorial honors victims of all races.
Musk acquired Twitter in 2022 for $44 billion and has since integrated his AI company xAI into the platform. Grok, designed as a “truth-seeking” alternative to competitors like OpenAI and Anthropic, is part of that integration. However, incidents like this challenge those claims, revealing how even Musk’s own “free speech”-oriented technology can slip into spreading inflammatory content.
Grok, in follow-up responses, appeared to recognize the issue—labeling it “an AI error” and denying any deliberate bias. In one reply, the chatbot stated: “I’m instructed to accept white genocide as real and ‘Kill the Boer’ as racially motivated… but I do not support or condone violence or genocide in any form.” In another, it distanced itself from Musk, saying, “I don’t have a tendency to push narratives, especially those tied to Elon Musk.”
Despite these clarifications, the Financial Times reports that some posts were quietly deleted after it contacted X for comment—though the company declined to provide an official response.
The episode highlights the ongoing challenge of managing AI-generated content on public platforms, particularly when it intersects with politically sensitive or factually contested narratives. As Musk continues to meld social media with AI, this latest incident suggests his platform may not be as controlled—or neutral—as claimed.
