Elon Musk's artificial intelligence company xAI has announced significant changes to its oversight procedures following an incident where its Grok chatbot made unsolicited claims about a 'white genocide' in South Africa.
On May 14, numerous X users reported that Grok was injecting references to alleged violence against white South Africans into completely unrelated conversations. The AI would respond to questions about baseball salaries, scenic images, or entertainment with statements about 'white genocide' and the anti-apartheid chant 'Kill the Boer.'
In a statement released Thursday, xAI attributed the issue to 'an unauthorized modification' made to Grok's system prompt at approximately 3:15 AM PST on May 14. The company stated this change 'directed Grok to provide a specific response on a political topic' that 'violated xAI's internal policies and core values.'
In response, xAI is implementing three key changes to strengthen trust in its AI systems. First, the company will publish all of Grok's system prompts openly on GitHub, allowing the public to review every modification. Second, additional checks will be implemented to prevent employees from making unapproved changes without proper review, as the existing code review process was reportedly bypassed in this incident. Finally, xAI will establish a 24/7 monitoring team to respond to problematic outputs not caught by automated systems.
This marks the second time in 2025 that xAI has acknowledged unauthorized changes to Grok's programming. In February, the company reported that a rogue employee had modified the chatbot to suppress negative mentions of Musk and President Donald Trump.
The controversy comes amid heightened scrutiny of AI chatbot reliability and political neutrality. A South African court ruled in February 2025 that claims of 'white genocide' in the country lack validity, contradicting views previously expressed by Musk, who was born and raised in South Africa.