Grok, Elon Musk’s chatbot, created a stir when it made baseless, racially-charged remarks about South Africa in random user posts.

Elon Musk’s AI chatbot, Grok, has caught eyes when responding to seemingly innocent posts on X with unrelated and controversial answers. Users said that they were shocked when the chatbot started speaking of the alleged violence against white South Africans — a subject that is associated with “white genocide” claims — even in the posts about sports, fish, or breathtaking

Even though Grok had no direct cue on race or South Africa, Grok kept on diverting discussions towards racially loaded narrations. In one case, it commented on a walking trail picture with the reference to South African farm attacks. As in another, it replied to a kitten video with a claim that refuted existence of evidence of white genocide but still relating it to statistics of general crime.

The glitch-like behavior has put chatbot credibility under the light. Experts say that chatbots sometimes “anchor” on a misinterpreted theme and cannot steer without guidance. Musk’s team has so far not provided an explanation for the confusing cross-talk, though some of Grok’s replies have apparently been deleted.

This episode revives controversies with regards AI ethics, chatbot training biases and the cleavage of political leanings within chatbot behaviors.