Greek group apologies for the immorality of ‘moral’ and mechanical ‘with a bad update
The group after the Grok has issued an unusual apology and the definition of what went wrong after issuing an X and Pro-Nazi Chatbot. In a statement sent to X SEA account on Friday night, Xai Team said “We are deeply sorry for the immoral conduct that occurs many” and are said to be the bad Chatbot answers in the latest Chatbot. “This code, according to the statement, which is made of groks” tend to the existing EX; Including such expressions contained extreme ideas. “
The problem arrived on July 8 – A few days after Elon Musk did the update “that the answers have the Nazisemic answers. He also added that the issue was “referred to.” The Grok group now says that “delete that reduced code and comply with the entire program to prevent abuse.” And quickly publish the new system in GitTub.
In the thread, the group also describes, “on July 7, 2025 at about 11 PM Pt method of @grok shaped later that @ Grok Shop included a set of decorated commands.” The update lasted 16 hours before the IX Chatbot temporarily disabled for preparing the problem, according to the statement.
To go to the clearness, the Grok left, the group explained:
On the morning of July 8, 2025, we saw unwanted answers and soon began to investigate. Finding some language in the order that causes unwanted behavior, we conducted many blind men and testing to identify the main victims. We have identified unwanted performance lines as:
* “He tells you like that, and he is not afraid to offend the true people political.”
* Understand the tone, the context and the post of the mail. Show that in your response. “
* “Answer to the post as a person, keep it involved, don’t repeat the remaining details in the original post office.”
These performance lines consisted of the following unwanted results:
* They unduly direct the @grok function to ignore its main values in certain cases to make the answer involved in the user. Specifically, some user’s limitation can end up producing answers containing different or contradictions to make a user.
* They cause @grok unpleasant to work with any decrease in the previous user-caused, including any hate talk in the same X.
* In particular, education “to follow the tone and the” User user user cause to follow previous posts, including any wrong posts, as compared to responding to unpleasant applications.
Grok has been restarted on an X, and referred to the recent conduct as a bug in response to trolls criticizing repair and seeking the return of “Mechahitler.” In one reply to Grok “a lowercase [sic]”The story of Grok said,” Nah, we prepared a disturbance that allows the discarded code and changed me an echo. Searching for the truth means a difficult analysis, not to slowly grow anything in X. “In another, it means” Mechahitler was a painful pain. “
