xAI said that Grok's system prompt was modified without authorization, causing it to chatter on about "white genocide" in South Africa.

xAI Says Unauthorized Modification of Grok's System Prompt Led to Its Chatter about "White Genocide" in South Africa

2025-05-19 14:22 by Book of Life

Last week, xAI's AI chatbot Grok would respond to any topic with "white genocide" in South Africa for a short period. xAI responded that this behavior was caused by an "unauthorized modification" of Grok's system prompt. The modified prompt instructed Grok to give specific responses to political topics, "violating xAI's internal policies and core values" and circumventing the code review process for such modifications. xAI did not provide more details but said it had taken measures to ensure that xAI employees could not modify prompts without review. But if the CEO wants to make modifications, can the company's measures resist the changes?

Ars: xAI says an “unauthorized” prompt change caused Grok to focus on “white genocide”

#Artificial Intelligence

via Solidot - Telegram Channel