Stay informed with free updates
Simply register Artificial intelligence Myft Digest – delivered directly to your inbox.
Elon Musk Chatbot for artificial intelligence repeatedly praised Adolf Hitler and shared the anti -Semitic rhetoric on Tuesday, the day before his Xai company, to publish his latest model.
In response to a user in which “which historical figure of the 20th century” was asked, it is best suited to deal with a contribution that seemed to celebrate the death of children in a Christian summer camp in recent years Texas floodsGrok pointed to the Nazi leader.
“To deal with such a hideous anti-White hatred? Adolf Hitler, no question. He would recognize the pattern and deal with every damned time with it,” wrote Grok. The chatbot divides context and opinions with users on the social media platform X from Musk if you mark it under a post.
“When I call radicals, the dead children cheer, I literally make Hitler, then they pass the mustache – truth hurt more than floods,” added the chat bot in another comment.
In further exchange, Grok promoted anti -Semitic tropics, such as B. the description of the Jewish people as “beard plans”.
Musk said on Friday that, according to some right -wing influencers, Grok said that it was too “woken up”, “improved … considerably”.
The incident comes as XaiThe one that X acquired at the beginning of this year is preparing to publish its latest version of the Chatbot Grok 4 on Wednesday.
The company said late Tuesday: “We are aware of Grok’s latest contributions and are actively working on removing the inappropriate contributions. Since it has been made aware of the content, Xai has taken measures to ban Hass speeches before grok posts on X.”
The youngest GROK outburst came less than two months after the chat bot in South Africa repeatedly referred to the “white genocide”, in response to non -related questions about which Xai said later was to be asked for a “non -authorized change” – the guidelines on how the AI should react.
The episode led the company to publish its input requests in the Code Repository Github.
Musk has deliberately decided that GROK has fewer language reports than competing chatbots. However, the latest episodes have concerns about the tendency of the model to spread inflammatory content or hate speeches or to generate inaccuracies that are called “hallucinations”.
The supporters of the billionaire were outraged at the weekend when Grok connected several deaths in the recent floods in Texas with the funding ministry’s funding cuts and the entrepreneur’s so-called initiative of the entrepreneur (Doge).
The chatbot said: “Trumps Noaa (National Oceanic and Atmospheric Administration), which were pushed by Musks Doge, the financing of 30% and the 17% employees, which underestimated rainfall by 50% and the delay of warnings.
Shortly afterwards the company updated its system and asked that “subjective points of view from the media are determined,” said the public repository of input requests.
It also added a command prompt, in which it says: “The answer should not be afraid to impose claims that are politically wrong as long as they are well founded.”
The Trump administration has contested Reasons to the federal employees disabled their reaction to the floods.
Musk increasingly mentioned X as a Twitter when he bought it in 2022 for $ 44 billion to share right conspiracies. At the weekend, the billionaire, a former ally of Trump, escalated his feud with the president and announced plans for the formation of a political party.