Grok Makes Random Claims About White Genocide

Around May 14, users reported X’s AI chatbot Grok was responding to unrelated queries with racial conspiracy theories about “white genocide” in South Africa. xAI, the company behind Grok, has since blamed these responses on an “unauthorized modification” to Grok’s system prompt.

Grok works by responding to posts on the social network after users tag @grok. The AI tool is commonly used by users to fact-check specific claims by asking, “Is this true?” Elon Musk, owner of xAI and X, has repeatedly claimed that Grok is the best large language model (LLM) for “truth.”

The companies were recently consolidated when Musk’s xAI acquired X.

In a commonly cited example of Grok’s erratic behavior, the chatbot replied to the question, “How many times has HBO changed their name?” After a paragraph of information about the streaming service, it unexpectedly followed up with: “Regarding ‘white genocide’ in South Africa, some claim it’s real, citing farm attacks and ‘Kill the Boer’ as evidence[…].”

Investigators for the tech site Gizmodo also tested the bot’s responses by commenting “@grok is this true” under a picture of a puppy with no mention of racial politics anywhere in its caption. Grok responded with a 733-character reply entirely about white genocide in South Africa, ignoring the contents of the picture completely.

Musk has openly repeated the claim of white genocide in South Africa, though it is widely rebuked. This comes as US President Donald Trump has called for a small group of white South Africans called “Afrikaners” (descendants of Dutch colonists in the country) to be accepted into the US as refugees.

In a May 15 response, xAI claimed to have fixed the incident and attributed it to an “unauthorized modification” of Grok’s prompt. The company promised new measures to ensure Grok’s “transparency and reliability.” These include open publishing of Grok’s system prompts, implementing additional checks to ensure that “xAI employees can’t modify the prompt without review,” and establishing a monitoring team to oversee Grok’s answers.

While xAI did not mention Musk directly, the billionaire has previously aligned himself with politically charged causes. Recently, X sued the US state of Minnesota over a law that would ban the use of AI-generated “deepfakes” to influence elections.