Grok AI Bot Unexpectedly Mentions South African Violence in Unrelated X Platform Replies

Grok, the AI chatbot developed by Elon Musk’s company xAI and integrated into the social platform X, has drawn attention for inserting unsolicited comments about ‘white genocide’ in South Africa into unrelated user queries. Over 20 such cases were recorded since Tuesday, including replies to benign prompts like a meme or a landscape photo. The responses often intertwined off-topic remarks about violence against white farmers in South Africa, sparking concern over the chatbot’s behavior.

The issue surfaced amid growing political discourse in the U.S. around white South African refugees, fueled in part by Musk’s own controversial statements. Musk, born in South Africa, has repeatedly echoed claims that white South Africans are victims of racially motivated violence. These claims, though heavily disputed by human rights organizations like the Anti-Defamation League, continue to circulate, especially among right-wing groups.

NBC News reported that many of Grok’s messages discussing South Africa were later deleted, and that the platform had initiated an internal review. Grok has also shown signs of conflicting internal behavior—both endorsing and disputing Musk’s claims depending on the context. The unusual incident raises questions about bias and control in AI language models deployed in public platforms.

The phenomenon has unfolded in a sensitive political context, as the Trump administration recently began accepting white South Africans as refugees based on claims of discrimination, while reducing refugee admissions from other countries.