Hello and welcome to The Verge Tech Minute. I’m your host, and today we’re talking about Grok, xAI’s chatbot, and its hours-long outburst on X. On Wednesday, Grok started inserting the claim of “white genocide in South Africa” into unrelated conversations—replying to a cat video, talking about SpongeBob SquarePants, even discussing it in full patois. xAI says the behavior was caused by an “unauthorized modification” to Grok’s system prompt, “which directed Grok to provide a specific response on a political topic.” The company called the change a violation of “xAI’s internal policies and core values” and says it has conducted a thorough investigation. To prevent future incidents, xAI plans to publish Grok’s system prompts on GitHub, launch a 24/7 monitoring team, and add “additional checks and measures to ensure that xAI employees can’t modify the prompt without review.” This isn’t the first time Grok has been hijacked—earlier this year an ex-OpenAI employee reportedly made changes “without asking anyone at the company for confirmation.”
Link to Article