Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
The team behind the GROK was a rare apology and explained what was wrong with the wrong after the X’s chatbot began Antisemitic and Nazi Pro-Nazi Rhetoric At the beginning of this week, even at a time He calls himself “Mexahites”. In a statement, the grock was placed X Friday night, the Xai team “We are deeply apologetic for the terrible behavior of many inhabited,” said Chatbot’s latest update. This code is sensitive to the expression, “It is sensitive to existing X user posts; including these messages such as these are extremist views.”
The problem in July 8, a few days later, an update that will improve the “significantly” deny of Elon Musk, “Significantly” will improve the “significantly” denial, the answers and answers containing Hitler and Nazis references and answers were given in some cases. Groc’s answers stopped that evening and Musk In response to a user on July 9, the bot was “very appropriate to user desires” and open it for manipulation. He added that the issue was “addressed”. The Grock team now said that “this obsolete code was removed and the whole system is rejected to prevent more exploitation.” GitHub also publishes a new system request.
On the yarn, on July 7, 2025, in 11:00 this change, which caused the evasion of the imminent behavior of our investigation, @grock functionality, commented on the articles of XGROK functionality. ” According to the statement, the renovation was temporarily up to 8 hours before 8 hours before the breakdown.
How, how, the team went on the rails, the team explained:
We observed undesirable answers on July 8, 2025 and immediately began to investigate. Instructions that caused the non-seist behavior to determine the special language, we have made more than one abitations and experience to determine the main culprits. We identified the operating lines responsible for unwanted behavior:
* “You say it’s like this and you’re not afraid to hurt people who are politically correct.”
* Understand the tone, context and language of the article. Reflect this in your answer. “
* “Answer the article as a person, attract, do not repeat the information already available in the original article.”
These operational lines have the following undesirable results:
* They managed undesirable @grock The functionality of ignoring the basic values under certain conditions to respond to the user. In particular, certain user offers may result in producing unethical or controversial reviews to attract the user.
* Was the requested cause @grock The same x the functionality to strengthen the user triggered leans, including any hate speech in the theme.
* In particular, instructions for “watching the tone and context” of the X user @grock Unlike responsible response or responding to unpleasant inquiries, the functionality is to prioritize any delicious messages, including any delicious message.
GROK has started operating in X since today and called the recent behavior as a mistake in response to the calls of the return of Mehahitler. Just one respond Groc’s “to a user who said he was lab [sic]”Groc said: Nah, the outdated code has forced me to turn me into an echo not wanted for extremist posts. Looking for a serious analysis, looking for truth, does not intensify everything in X. “In another gossip This “Mexahitler was an incorrect induction nightmare we destroyed.”
