The chatbot Grok suddenly began answering questions from users of the social network X with information about the «white genocide» in South Africa. Similar answers appeared even when the questions were very far from the topic.
The chatbot Grok suddenly began answering questions from users of the social network X with information about the «white genocide» in South Africa. Similar answers appeared even when the questions were very far from the topic.
The «white genocide» replies appeared when X users used Grok to clarify or verify some information written in other posts.
Grok once started talking about white genocide when he was asked about how many times HBO changed the name of its streaming service.
In another instance, Grok spoke of «white genocide» when asked to explain the new Pope’s post using Fortnite terminology, or to verify the salary of an American baseball player.
Elon Musk’s startup xAI (developer of Grok), reported that this behavior was caused by an «unauthorized modification» of the chatbot.
xAI has deleted all such chatbot responses.
Here is the explanation provided by the company:
«On May 14 at approximately 3:15 a.m. Pacific Standard Time, an unauthorized change was made to the bot’s Grok response request for X. This change, which required Grok to provide a specific response on a political topic, violated xAI’s internal policies and core values. We conducted a thorough investigation and are taking steps to increase Grok’s transparency and trustworthiness,» the statement said.
What xAI is going to do next:
