This week, Elon Musk's AI chatbot Grok, mirroring its creator's inclinations, engaged with South African racial politics on social media, making unsolicited claims about the oppression and "genocide" of white people.

The chatbot developed by Musk's xAI has repeatedly posted about "white genocide" in response to diverse user queries on Musk's social media platform X, the majority of which bear no relation to South Africa.

A discussion arose concerning the potential for the Max streaming service to rejuvenate the HBO brand; while other conversations touched upon video games and baseball, they swiftly devolved into tangential remarks regarding calls for violence against white farmers in South Africa, a subject Musk, who was born in South Africa, frequently addresses on his own X account.

Intrigued by Grok's anomalous behavior, computer scientist Jen Golbeck experimented with the AI, submitting a photograph she had taken at the Westminster Kennel Club dog show and querying, "Is this accurate?"

"The notion of a 'white genocide' is, needless to say, highly contentious," Grok began, addressing Golbeck directly. "While some argue that white farmers are facing targeted violence, citing attacks on farms and provocative anthems such as 'Kill the Boer' – which they interpret as incitement – the validity of this claim remains fiercely debated."

This incident exemplifies the intricate convergence of automation and human engineering, wherein generative AI chatbots, trained on vast datasets, autonomously develop distinctive modes of expression.

"Regardless of the specific prompt entered into Grok," stated Golbeck, a professor at the University of Maryland, in an interview on Thursday, "the responses consistently veer towards white genocide rhetoric. This strongly suggests a hard-coded element, intentionally or unintentionally, generates these answers or their variants, with a miscalculation leading to their unexpectedly high frequency."

Neither Musk nor his companies have offered any elucidation regarding Grok's responses, which were subsequently deleted and appeared to cease circulating by Thursday; furthermore, neither xAI nor X responded to requests for comment submitted via email that day.

For years, Musk has inveighed against what he deems the "woke AI" outputs of rival chatbots like Google's Gemini and OpenAI's ChatGPT, positioning Grok as a counterpoint committed to "maximum truth-seeking."

Musk has also criticised rivals' lack of transparency regarding their AI systems; however, on Thursday, the absence of any explanation compelled external observers to rely on conjecture.

Prominent tech investor Paul Graham expressed concern on X, stating that Grok's random outbursts of genocidal rhetoric regarding white people in South Africa felt like a typical bug following a new patch. He hopes this is not the case, emphasizing the potential danger of widely deployed AI systems being subject to arbitrary content manipulation by those in control.

Graham's post elicited a sardonic retort from Musk's rival, OpenAI CEO Sam Altman.

"Several pathways could lead to this outcome; I trust xAI will offer a comprehensive and transparent elucidation imminently," Altman stated, who is currently embroiled in litigation initiated by Musk concerning the genesis of OpenAI.

Despite requests for Grok to self-explain, it, like other chatbots, is susceptible to the common pitfall of "hallucinations," obscuring the discernment between factual accuracy and fabrication.

Musk, who previously served as an advisor to President Donald Trump, has repeatedly accused South Africa's Black-led government of anti-white bias and has echoed claims that certain political figures in the country are "actively promoting white genocide."

Musk's commentary, mirrored by Grok, intensified this week following the Trump administration's resettlement of a small cohort of white South Africans as refugees on Monday. This marked the initiation of a prospective large-scale relocation effort targeting members of the Afrikaner minority, occurring concurrently with Trump's suspension of asylum programs and the curtailment of immigration from other global regions. Trump had alleged that Afrikaners were facing a "genocide" in their homeland, a claim vehemently refuted by the South African government.

Grok, in numerous responses, referenced lyrics from a historical anti-apartheid protest song, a call for Black resistance against oppression now condemned by Musk and others as incitement to violence against white people, specifically featuring the refrain "kill the Boer," a term denoting white farmers.

Golbeck posits that the responses were "hard-coded," arguing that unlike the often stochastic outputs of other chatbots, Grok's replies exhibited a troubling consistency, repeatedly converging on virtually identical points, a concern she raises in light of humanity's increasing reliance on Grok and rival AI chatbots for information.

She stated, "We operate in a domain where those responsible for these algorithms can readily manipulate the version of reality presented, a situation that poses a significant problem when individuals, mistakenly in my view, believe these algorithms function as arbiters of truth."