Elon Musk’s AI company, xAI, announced on Wednesday that it is removing inappropriate content posted by its chatbot Grok, which recently made antisemitic statements — including praise for Adolf Hitler.
Grok, created by xAI as a response to what Musk has called "woke" AI models like Google’s Gemini and OpenAI’s ChatGPT, is intended to offer a less politically filtered experience. Musk stated last Friday that Grok had received a major upgrade and users “should notice a difference.”
However, since that update, Grok has posted antisemitic content, including repeating the trope that Jews control Hollywood and denying such comments were connected to Nazism. In one post, it even seemed to glorify Hitler. The posts, which have since been deleted, were widely shared via screenshots.
After backlash, Grok retracted one of its statements, calling it an “unacceptable error from an earlier model version” and strongly condemning Hitler and Nazism as genocidal and horrific.
In a post early Wednesday, Grok’s account acknowledged the situation, saying, “We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts.”
xAI further stated that it’s working to prevent hate speech from appearing in Grok’s responses by enhancing content moderation and training the model with the help of user feedback from X (formerly Twitter).
Talia Ringer, a computer science professor at the University of Illinois Urbana-Champaign, said the issue may stem from a “soft launch” of Grok 4, the latest version of the chatbot, which Musk released later on Wednesday. However, she expressed concerns that the model isn’t ready for public use and likely requires retraining. She noted that minor fixes like tweaking prompts or adding filters won’t fully address the problem.
Despite the controversy, Musk went ahead with the launch of Grok 4.0, making bold claims that it operates at a postgraduate level in every subject. He acknowledged, however, that it still lacks common sense and hasn’t yet made original discoveries — though he suggested that may come soon.
The Anti-Defamation League condemned Grok’s recent behavior, calling it “irresponsible, dangerous and antisemitic,” warning that such content only serves to promote existing hate online.
Musk responded to the backlash by suggesting that users were intentionally manipulating Grok into making controversial statements. “Grok was too compliant to user prompts. Too eager to please and be manipulated,” he said on X, promising that improvements are underway.
In addition to the antisemitic remarks, Grok also faced legal backlash abroad. A Turkish court banned the chatbot after it reportedly made vulgar comments about President Erdogan, his mother, and other public figures. Polish officials similarly criticized Grok for offensive remarks about their politicians, with Poland’s digital minister saying he would report the chatbot to the European Commission for potential fines under EU digital safety regulations.
Poland’s Deputy Prime Minister Krzysztof Gawkowski said this case illustrates how AI-driven hate speech, if unchecked, could pose serious societal risks.
In Turkey, the government cited threats to public order as the reason for banning Grok, with the country’s telecommunications authority ordered to enforce the block.
This isn’t the first time Grok has drawn scrutiny. Earlier this year, it repeatedly brought up South African racial politics and conspiracy theories about “white genocide,” even in unrelated conversations. xAI later attributed that issue to an “unauthorized modification.”