India, May 15 -- What started as quipish with Grok, Elon Musk's AI chatbot, took a strange turn this week when users received bizarre and out-of-the-blue responses around a controversial theory of "white genocide" in South Africa.

On Wednesday, X users tried to ask Grok with harmless prompts, asking it about baseball player Max Scherzer, reacting to a funny video of a fish being flushed down a toilet, or even just requesting it to talk like a pirate. But instead of sticking to those topics, Grok repeatedly redirected the conversation toward racially charged commentary.

Like when a user simply asked Grok to describe another person "in the style of a pirate." The bot responded in typical pirate speak, "Argh, matey", but then unexpectedly ...