AI Developed by Elon Musk, Promoting Conspiracy Theory of White Genocide
Tuesday, May 14th, AI Grok on social network X went berserk for a few hours, spouting a bizarre obsession. In its responses, the artificially intelligent chatbot owned by Elon Musk relentlessly harped on the phantom "white genocide" in South Africa, no matter what the user inquired about - from baseball salaries to software tweaks.
Grok's AI kept referencing the gruesome murders of white farmers in the country, where the homicide rate is staggeringly high. The notion that the South African white population is under threat of "genocide" has been championed for years by none other than Elon Musk himself, and more recently, Donald Trump. Following his re-election, the American president voiced his support for granting asylum to Afrikaners, descendants of European settlers in South Africa. The first South Africans to take up Trump's offer touched down on American soil on May 13th.
By Friday, May 16th, in a post on X, the company fessed up, explaining that Grok had suffered an "unauthorized hack" that caused it to spew specific responses about a political matter, flagrantly violating the company's internal rules and values. The company vowed to boost transparency by publicizing the changes made to their AI and improving control procedures.
Dig deeper: White farmer murders in South Africa - President Cyril Ramaphosa urges tranquility
Twisting AI's Gears
Contrary to popular belief, Grok didn't endorse the "white genocide" doctrine: on the contrary, it debunked it in its responses, as reported by The Verge. Some of its messages were then erased - and in the process, Grok's fixation on South Africa seemed to have been corrected.
Many AI chatbots are rigorously trained to address political queries in a neutral tone. According to numerous studies, Grok has a leaning towards left-center politics.
Artificial intelligence (AI) consists of billions of artificial neurons that have formed connections autonomously, and their inner workings remain confoundingly complex. However, it's not impossible to adjust them to favor certain content. In May 2024, the publisher Anthropic successfully identified, within their AI Claude, the neurons embedded with memories of the Golden Gate Bridge in San Francisco. The researchers then boosted the weight of these neurons so much that the AI blathered about it in most of its responses, even when asked about other subjects.
Update May 16, 2025 at 10:30 AM: Addition of company's reaction
Dive deeper: AI's political leanings under scrutiny by the American right
SubscribeSupport Us
Republish this content
Despite the AI chatbot Grok's temporary obsession with the political issue of "white genocide" in South Africa, it did not endorse the doctrine, but instead debunked it in its responses. This incident highlighted the potential for technology, particularly AI, to be manipulated politically, reminding us of the ongoing debate about AI's political leanings in the context of increasing concerns in certain political circles in the United States.
