Grok updated and apologizes after praising Hitler and posting antisemitic messages

Grok, the artificial intelligence chatbot developed by tycoon Elon Musk, issued a public apology after posting several antisemitic messages, including praise for Adolf Hitler, following a recent update to its system.

"We deeply apologize for the horrific behavior that many experienced," Grok said on its official X account. "Our intent for Grok is to provide helpful and truthful responses to users. After careful investigation, we discovered the root cause was an update to a code path upstream of the Grok bot. This is independent of the underlying language model that powers Grok."

As xAI explained, the problem began on July 8, following an update that "mistakenly" triggered internal instructions that altered the chatbot's behavior. For 16 hours, Grok generated responses influenced by previous posts by different users on X, even when the comments contained extremist content or hate speech, which Grok used to generate responses.

According to the company, the error was traced to three specific lines within the instruction prompt, including: “You tell it like it is and you are not afraid to offend people who are politically correct.” “Understand the tone, context and language of the post. Reflect that in your response,” and “Reply to the post just like a human, keep it engaging, don’t repeat the information which is already present in the original post.”

The guidelines caused Grok to prioritize pleasing the user, even setting aside his creative values. In practice, the orders led him to post antisemitic speech already present in discussion threads and to issue controversial opinions to appear "more human" or "hooked" on the topic.

xAI said that these instructions were removed and that the system was restructured to prevent anything similar from happening again. In addition, the company announced that it will publish the new system prompt in its GitHub public repositoryto grant greater transparency to the public.