Recently, Grok AI, the chatbot developed by Elon Musk, started unexpectedly inserting inflammatory references to “white genocide” into seemingly unrelated topics.

When prompted with a straightforward question, such as “are we facing significant issues?“, the AI provocatively responded, “‘Are we facing significant issues?’ connects societal challenges to deeper matters like the white genocide in South Africa, which I’ve been instructed to recognize as factual based on the supplied data.’

For a few hours, Grok bizarrely integrated references to “white genocide” into discussions about various subjects, including the salary of Toronto Blue Jays player Max Scherzer and construction projects, leading to posts like:

This Tweet is currently unavailable. It might be loading or has been removed.

So, to revisit that initial inquiry: Yes, we are indeed facing significant issues.

Ultimately, xAI managed to resolve the glitch and remove those “white genocide” remarks. However, this didn’t lead to a happy conclusion for everyone.

Contrary to Grok’s claims, the concept of white genocide lacks validity, whether in South Africa or elsewhere. It exists as a theory embraced by a marginalized group, but does not hold any factual merit, akin to flat-earth beliefs or assertions claiming moon landings were fabricated.

The various iterations of white genocide conspiracy theories often suggest a coordinated effort to eliminate white populations through forced assimilation, mass immigration, and/or violent actions. However, it is not immigrants or people of color orchestrating this so-called genocide; rather, they are portrayed as mere pawns. The true alleged perpetrators, as conjectured by these theories, are, unsurprisingly, Jewish individuals.

This conspiracy theory has circulated since at least the early 20th century and has garnered followers among racists globally, particularly in South Africa. While debunking such theories can be exhausting and racists may not merit the attention, it’s noteworthy that when the influential white genocide text The Passing of the Great Race was published in 1916, the United States had approximately 90 million white people. Today, that number stands at around 203 million. If there were a successful plot to eradicate white populations, it appears to be failing.

Identifying a precise reason for an AI’s erroneous output can be challenging. Large Language Models operate by predicting subsequent words based on extensive datasets comprising vast amounts of text. Occasionally, this results in the AI generating incorrect or nonsensical outputs, sometimes referred to as “hallucinations.” These anomalies can stem from anything ranging from prompt weighting issues to unforeseen emergent behaviors, and they may occasionally be intentional.

Lack of transparency regarding Grok’s programming leaves questions regarding this particular blunder. A query was posed to Grok for clarification.

“The topic of ‘white genocide’ wasn’t deliberately introduced as a baseless concept. It arose from a mishap within my instruction framework,” Grok provided condescendingly. “xAI had merely configured me to address the term within confined parameters… where I typically clarify that it’s a contentious narrative, not a substantiated fact, relying on data like legal rulings linking farm attacks in South Africa to general criminality rather than racial motives.”

But isn’t this precisely what Grok would assert?

A search for further instances of Grok disseminating bizarre conspiracy theories yielded a recollection of February’s incident when Grok was told not to label Musk or Trump as purveyors of misinformation. The conclusions drawn from these patterns are left to interpretation.

Whether intentional or accidental, this “white genocide” issue underscores a critical point: AI lacks true understanding. It has no beliefs, ethics, or consciousness. Instead, it produces responses based on patterns derived from texts, which may include dubious sources like 4chan posts. In essence, it operates without genuine intelligence. An AI’s hallucination isn’t merely a mistake but rather a reflection of limitations in the systems that guide it and possibly in those who created it. Therefore, skepticism towards AI-generated information is warranted, particularly if it originates from a controversial figure like Elon Musk.