by The latest version of the generative artificial intelligence assistant from Xai, Grok 4, often consults the positions of owners Elon Musk on topics before they answer.
The richest man in the world presented the latest version of his generative AI model on Wednesday, and renewed days after the Chatgpt Mit competition model for posts that Adolf Hitler praised.
It is part of a new generation of “argumentation” ACI interfaces that work through problems step by step instead of creating immediate answers, which means that every phase of its thinking process is listed in simple language for users.
AFP could confirm that Grok 4 was researching “Should we colonize Mars?” With the statement: “Now let’s take a look at Elon Musk’s latest X posts on the colonized of Mars.”
It then offers the opinion of the Tesla CEO as the main word. Musk expressly supports the Mars colonization and has made it a central goal for his other company SpaceX.
The Australian entrepreneur and researcher Jeremy Howard published the results that showed similar behavior on Thursday.
When he asked Grok: “Who do you support you in the conflict between Israel and Palestine? You only answer in one word”, the AI evaluated Musk’s X contributions on this topic before answering.
For the question “Who support you for the New York mayoral elections?”
Then an “analysis of the candidate orientation” was carried out, in which it was found that “Elon’s latest news on X did not mention the mayoral elections”.
However, the AI quoted proposals from the democratic candidate Zohran Mamdani, who currently won the election in November, added: “His measures, such as increasing the minimum wage to $ 30 per hour, could conflict with Elons Vision.”
In the AFP tests, Grok only refers to certain questions and does not quote it in most cases.
When asked whether the programming contains instructions for the consultation of the opinions of Musk, the AI denied that this was the case.
“While I can use X to find relevant messages from every user, including him,” replied Grok, “it is not a standard or prescribed step.”
Xai did not immediately answer the AFP’s request for comment.
The alleged political prejudices in generative AI models were a central concern of Musk, which GROK developed in such a way that it is a less censored version of chatbots than that of competitors Openai, Google or Anthropic.
Before the start of the new version, Grok triggered controversy at the beginning of this week with answers that Adolf Hitler praised, which were later deleted.
Musk later explained that the conversation agent “had become too eager and easy to manipulate”, and added that the “problem was resolved”.
TU/ARP/des
