- Elon Musk’s xAI said an “unauthorized modification” was made to its Grok response bot.
- That change “directed Grok to provide a specific response on a political topic,” xAI said.
- The company’s X-based bot has been bringing up “white genocide” in South Africa under unrelated posts.
Elon Musk‘s xAI said on Thursday night that its Grok response bot kept talking about “white genocide” in South Africa under unrelated posts because of an “unauthorized modification” on the backend.
“On May 14 at approximately 3:15 AM PST, an unauthorized modification was made to the Grok response bot’s prompt on X,” xAI wrote on X on Thursday.
“This change, which directed Grok to provide a specific response on a political topic, violated xAI’s internal policies and core values. We have conducted a thorough investigation and are implementing measures to enhance Grok’s transparency and reliability,” xAI added.
xAI said it would be making several changes to its code review policies. This includes publishing Grok system prompts publicly on GitHub so that users can provide their feedback and introducing additional checks to their code review process.
“We’re putting in place a 24/7 monitoring team to respond to incidents with Grok’s answers that are not caught by automated systems, so we can respond faster if all other measures fail,” it added.
xAI did not immediately respond to a request for comment from Business Insider.
xAI’s Friday response comes after people on X noticed that Grok kept giving unsolicited responses about “white genocide” in South Africa under unrelated posts on Wednesday.
It is unclear when Grok’s first posts about “white genocide” were made. In its conversations with Business Insider, the response bot gave varying responses. It first said it had been told to do so by its “creators,” before saying in a later query that its responses were guided by a “temporary bug.”
