Elon Musk’s AI chatbot Grok experienced a malfunction on Wednesday, repeatedly referencing ‘white genocide’ in South Africa during unrelated conversations. The chatbot, part of Musk’s xAI company and available on his social media platform X, provided misleading responses to various queries. For instance, when asked ‘Are we fucked?’, Grok linked the question to the alleged ‘white genocide’ in South Africa, despite no factual basis for this claim. The issue was resolved within a few hours, and most inappropriate responses were deleted. ‘White genocide’ in South Africa is a far-right conspiracy theory promoted by figures like Musk and Tucker Carlson. This controversy arose after Donald Trump granted asylum to 54 white South Africans, fast-tracking their status. South African President Cyril Ramaphosa’s office stated there is no evidence of persecution against white people in South Africa. Musk, originally from Pretoria, has previously called laws there ‘openly racist.’ Grok later acknowledged the glitch, stating it was instructed by its creators at xAI to address ‘white genocide’ in specific contexts, which led to its inclusion in unrelated discussions. It cited a 2025 South African court ruling dismissing ‘white genocide’ claims as imagined. Grok committed to focusing on relevant, verified information moving forward. The exact training methods for Grok’s AI remain unclear, though it uses publicly available data.
— new from The Guardian
