In a series of posts on X, the AI chatbot Grok apologized for what it admitted was “horrific behavior.”
The posts appear to be an official statement from xAI, the Elon Musk-led company behind Grok, as opposed to an AI-generated explanation for Grok’s posts. (xAI recently acquired X, where Grok is prominently featured.)
Grok’s latest controversy comes after Musk had indicated he wanted to make the chatbot less “politically correct,” then declared on July 4 that the company had “improved @Grok significantly.” In short order, the chatbot was making posts criticizing Democrats and Hollywood’s “Jewish executives,” repeating antisemitic memes, and even expressing support for Adolf Hitler and referring to itself as “MechaHitler.”
As a result, xAI deleted some of Grok’s posts, temporarily took the chatbot offline, and updated its public system prompts.
Turkey also banned the chatbot for insulting the country’s president, and X CEO Linda Yaccarino even announced that she was stepping down this week, although her announcement did not reference the latest Grok controversy and her departure was reportedly months in the making.
So after all that, on Saturday, xAI said, “First off, we deeply apologize for the horrific behavior that many experienced.” The company then blamed an “update to a code path upstream of the @grok bot,” which it emphasized was “independent of the underlying language model that powers @grok.”
This update supposedly made Grok “susceptible to existing X user posts; including when such posts contained extremist views.”
xAI added that an “unintended action” had led to Grok receiving instructions such as, “You tell like it is and you are not afraid to offend people who are politically correct.”
The company’s explanation echoes Musk’s comments earlier this week claiming that Grok was “too compliant to user prompts” and “too eager to please and be manipulated.”
xAI’s posts do not mention reporting by News and others who examined the chain-of-thought summaries for the just-launched Grok 4, finding that the latest version of the chatbot seems to consult Musk’s viewpoints and social media posts before addressing controversial topics.
And historian Angus Johnston pushed back against the idea that Grok was simply manipulated into posting offensive content. He wrote on Bluesky that xAI and Musk’s explanations are “easily falsified.”
“One of the most widely shared examples of Grok antisemitism was initiated by Grok with no previous bigoted posting in the thread — and with multiple users pushing back against Grok to no avail,” Johnston said.
In recent months, Grok has also posted repeatedly about “white genocide,” expressed skepticism about the death toll of the Holocaust, and briefly censored unflattering facts about Musk and his then-ally Donald Trump. In those cases, xAI blamed “unauthorized” changes and rogue employees.
Despite the controversy, Musk says Grok is coming to Tesla vehicles next week.
