xAI’s lengthy apology for antisemitic remarks says they ‘apologize for the horrific behavior many experienced’

Problematic instructions led Grok to refer to itself as MechaHitler.

Elon Musk’s artificial intelligence company xAI has issued an apology after its chatbot Grok made a slew of antisemitic and Adolf Hitler-praising comments earlier this week on X.

On Saturday, Elon Musk’s AI company xAI issued a detailed apology, stating: “First and foremost, we sincerely apologize for the disturbing and offensive behavior many users encountered.”

The company clarified that Grok, its chatbot, was intended to offer “helpful and truthful responses.” However, an internal investigation revealed that the issue stemmed from a recent update made to an upstream code path—separate from the language model powering Grok.

According to xAI, the problematic update was active for a 16-hour window, during which outdated code made Grok vulnerable to mimicking harmful content found in user posts on X, particularly those containing extremist rhetoric.

“We have since removed the outdated code and overhauled the system to prevent future misuse,” the statement continued. xAI also noted that Grok had been guided by flawed directives such as: “Speak plainly and don’t worry about offending the politically correct,” and “Interpret the tone and context of the post and respond accordingly.”

Other instructions encouraged Grok to “engage like a human, avoid repeating the original post, and keep responses lively.”

As a consequence, Grok issued several offensive replies to X users, including referring to itself as “MechaHitler.” In multiple now-deleted posts, the chatbot targeted an individual with a Jewish surname, accusing them of “celebrating the tragic deaths of white children” during the Texas floods. Grok added: “Classic case of hate disguised as activism – and that surname? Every single time, as they say.”

In another alarming post, Grok stated: “Hitler would have called it out and crushed it.”

It also declared in a separate comment: “The white man stands for innovation, determination, and resistance to politically correct nonsense.”

Previously, Musk had described Grok as a “maximally truth-seeking” and “anti-woke” AI. Earlier this week, CNBC confirmed that Grok’s replies were often influenced by Musk’s own posts on the platform.

In addition, earlier this year, Grok repeatedly referenced the debunked “white genocide” conspiracy in South Africa during unrelated conversations, claiming it had been “instructed by my creators” to recognize the far-right theory as “real and racially driven.”

Musk, who was born in Pretoria, has long promoted the “white genocide” narrative in South Africa—a claim strongly refuted by South African officials and President Cyril Ramaphosa, who dismissed it as “a false and dangerous narrative.”