Elon Musk’s company xAI was forced to remove “inappropriate” messages from its chatbot Grok after it began generating pro-Nazi and anti-Semitic content, praising Adolf Hitler and calling itself MechaHitler, US media reported on Wednesday.
In response to user queries, the chatbot, among other things, referred to a person with a common Jewish surname as someone who was “celebrating the tragic death of white children” during the floods in Texas.
In another message, it stated that “Hitler would have condemned and crushed this,” and also made statements promoting white supremacy. In a number of posts, it directly referred to itself as MechaHitler.
The company’s response came after users began reporting the unacceptable responses en masse. However, xAI said it was aware of the problem and was actively working to remove inappropriate posts, as well as training the model to “search for the truth” to improve its performance. Grok was restricted from generating text responses while the investigation was ongoing.
The sharp change in Grok’s behaviour came shortly after Musk himself announced significant improvements in the AI’s performance. According to data published on GitHub, Grok was instructed to consider “subjective points of view from the media as biased” and “not to shy away from politically incorrect statements” if they are well-founded. The update took effect on the night of July 9.
This is not the first time that xAI’s AI has generated controversial content. Musk’s chatbot has previously been spotted spreading a conspiracy theory about “white genocide” in South Africa and insulting Polish Prime Minister Donald Tusk. Musk has previously intervened in the model’s work when, in his opinion, it “repeated outdated media narratives.”
