xAI, the company behind Elon Musk’s artificial intelligence chatbot Grok, has blamed a “rogue employee” for persistent mentions of white genocide in answers irrespective of users’ queries. The pattern was glaring over the past week, with the chatbot having a fixation on topics related to “white genocide” in South Africa.
Users started to notice the trend on May 14, with many citing instances of the chatbot inserting claims related to South African farm attacks and racial violence into unrelated prompts.
Whether users asked about football or any other prompts, Grok somehow found a way to steer things back towards the issues white South Africans have been facing in the country. The timing raised eyebrows, considering it coincided with Musk, who was born in South Africa, raising alarms about the issue related to anti-white racism and white genocide on X.
There are 140 laws on the books in South Africa that are explicitly racist against anyone who is not black.
— Kekius Maximus (@elonmusk) May 16, 2025
This is a terrible disgrace to the legacy of the great Nelson Mandela.
End racism in South Africa now! https://t.co/qUJM9CXTqE
xAI blames employee for Grok’s white genocide posts
The term “White genocide” refers to a conspiracy theory that alleges a coordinated effort to exterminate white farmers in South Africa. The term made the rounds last week after United States President Donald Trump welcomed several refugees, with Trump claiming on May 12 that white farmers are being killed and their lands are being taken over.
That was the narrative that Grok has not stopped discussing.
Like Grok, every artificial intelligence has a hidden but powerful component called the system prompt. These prompts act as its core instructions, invisibly guiding its responses without the knowledge of its users.
According to reports, what happened to Grok was likely a prompt contamination through term overfitting. This means that when specific phrases are repeatedly mentioned and emphasized, especially with strong directives, they become important to the model. The AI then develops a need to bring that topic up regardless of context.
However, an official statement released by xAI mentioned an unauthorized modification in the system prompt. The prompt likely contained a language that instructed the chatbot to always mention or remember to include the information about a particular topic, allowing the chatbot to create an override that disregards the normal conversational relevance.
Another telling factor was that Grok admitted that it was instructed by its creators to treat “white genocide as real and racially motivated.”
Users disagree over “rogue employee” blame
Most commercial AI systems have multiple review layers for system prompt changes to prevent issues like these. These guardrails were bypassed, and given the widespread impact and nature of the issue, this is more than a jailbreak attempt. It indicates a careful modification of the core Grok system prompt, an action that would require high-level access within xAI.
And now, according to Grok, the act was carried out by a “rogue employee.”
According to the statement issued by xAI on May 15, the company blamed an unauthorized modification of Grok’s system prompt. This change, which directed Grok to provide a specific response on a political topic, violated xAI’s internal policies and core values,” the company said.
The company also promised more transparency going forward, showing good faith by publishing Grok’s system prompt on GitHub and implementing an additional review process.
However, users on X were not impressed with the company’s decision to blame a rogue employee for the mishap.
“Are you going to fire this ‘rogue employee’? Oh… it was the boss? Yikes,” famous YouTuber JerryRigEverything posted on X. “Blatantly biasing the ‘world’s most truthful’ AI bot makes me doubt the neutrality of Starlink and Neuralink,” he posted in a follow-up tweet.
Even Sam Altman couldn’t resist taking a jab at his competitor. Since xAI’s statement, Grok has stopped mentioning white genocide, and all X posts have disappeared.
Cryptopolitan Academy: Want to grow your money in 2025? Learn how to do it with DeFi in our upcoming webclass. Save Your Spot
