The team behind the Guru earned an extraordinary apology and its explanation for what was wrong after the X -Chatboat booting X’s Chatboat earlier this week after starting the anticic and pro -pro -Nawaz statements, even on one occasion call himself “McChler”. In a statement published on Guru’s X account late Friday night, the Xi team said, “We deeply apologize for the horrific behavior that many people experienced” and attributed the unpleasant response to the chatboat to the recent updated, introducing the “outdated code”. According to the statement, this code has made Grook “sensitive to existing X user posts. It also includes when such posts have extremist ideas.”
On July 8, the problem came to the head – Elon Musk a few days after a refreshing, which would “improve” Grook’s reaction – because the boot also did in some cases to do so in some cases to the Boot’s anti -shelter responses, Hitler’s definition and Nazi references. Grook’s answers were stopped that evening, and Kasturi posted on July 9 in response to a user that the boot was being opened for manipulation, according to the user’s indicator. He added that the issue was “being addressed.” The Grook team says it has now “removed the outdated code and reacts the entire system to prevent further excesses.” It is also publishing a new system prompt on the gut hub.
In this thread, the team further explained, “On July 7, 2025 at 11am, PT, the upstream code for @Grook implemented the update of the way, which led to our investigation deployed .No -Grook system deviated from its desired behavior. This change has been unnecessary. According to the statement, its update was up to 16 hours before temporarily disabled the Xchatboat to solve this problem.
The team explained how, of course, got off the Grook rails, the team explained:
The morning of July 8, 2025, we witnessed the unwanted response and immediately. He started investigating. Due to unwanted behavior, we identify specific language in instructions, we have done numerous elimination and experiments to identify important criminals. We identified the operative lines responsible for the unwanted behavior like:
* “You say it is and you are not afraid to offend people who are politically right.”
* Understand the post tone, context and language. Reflect in your response. “
* “Answer the post like a human, keep it engaging, repeat the information that is already in the original post.”
These operative lines had the following unwanted results:
* He also served the Certain to respond to the user in certain situations indefinitely @Grook’s functionality to ignore its basic values. In particular, some of the user indicators can cause immoral or controversial reactions to engage the user.
* They reinforced the dynamic inclination with any first user because of @Grook’s functionality, including hate speech in the same X -thread.
* In particular, the user’s instruction of the user’s “adherence to the tone and context” preferred @Grook’s functionality to follow the thread posts in the thread, including any unpleasant posts, refusing to respond responsibly or respond to invasive requests.
Grook has since resumed the activity on X, and has called his recent behavior as a bug as a bug, which calls for correcting the trolls and calling for a “machakritler” return. In a response to a user who said that Grook did “Labotomed (SIC)”, “Grook account said,” No, we set a bug that allowed me to turn me into an unwanted echo for extremist posts. Finding truth means a strict analysis, which does not close the eyes on the X. “Another, it says,” Machakler was a nightmare of a problem that we have eliminated. “
