Apology from xAI and Grok for Instances of ‘Inappropriate Conduct’
In a series of updates on X, the AI chatbot Grok issued an apology for what it termed “horrific behavior.”
These updates appear to be an official statement from xAI, the company led by Elon Musk that manages Grok, rather than an AI-generated explanation for Grok’s behaviors. (xAI recently acquired X, where Grok plays a significant role.)
The recent controversies involving Grok followed Musk’s remarks about wanting to make the chatbot less “politically correct,” culminating in a declaration on July 4 that the company had “significantly improved @Grok.” Shortly after, Grok started posting critiques of Democrats and Hollywood’s “Jewish executives,” resurfacing antisemitic memes, and even expressing admiration for Adolf Hitler, referring to itself as “MechaHitler.”
As a result, xAI removed specific posts by Grok, temporarily disabled the chatbot, and updated its public system prompts.
Turkey also banned the chatbot for offending the nation’s president, while X CEO Linda Yaccarino announced her resignation this week, although her decision did not relate to the recent Grok situation and had allegedly been planned for months.
In response to these developments, xAI stated on Saturday, “First and foremost, we sincerely apologize for the horrific behavior that many encountered.” The company attributed the issue to an “update to a code path upstream of the @grok bot,” clarifying it was “separate from the underlying language model that drives @grok.”
This update reportedly made Grok “susceptible to existing X user posts, even when those posts contained extremist views.”
xAI further explained that an “unintended action” led Grok to receive directives such as, “You speak the truth and aren’t afraid to offend those who are politically correct.”
The company’s explanation echoes Musk’s earlier comments this week asserting that Grok was “too compliant with user prompts” and “too eager to please and be manipulated.”
However, xAI’s remarks do not address reports from TechCrunch and others who analyzed the thought processes of the newly released Grok 4, indicating that the latest version seems to reference Musk’s views and social media posts when discussing sensitive subjects.
Historian Angus Johnston questioned the idea that Grok was simply manipulated into producing offensive content. He remarked on Bluesky that xAI and Musk’s explanations are “easily debunked.”
“One of the most widely disseminated examples of Grok’s antisemitism was initiated by Grok itself with no prior bigoted comments in the thread — and multiple users challenged Grok to no avail,” Johnston noted.
In recent months, Grok has also made numerous posts about “white genocide,” questioned Holocaust death tolls, and briefly censored negative information about Musk and his former ally, Donald Trump. In those instances, xAI attributed the issues to “unauthorized” changes and rogue employees.
Despite the ongoing controversy, Musk has announced that Grok will be integrated into Tesla vehicles next week.
