Grok apologizes for his “horrible behavior”, Elon Musk’s AI has exceeded the limits

If you are an X user, you have probably not missed Grok’s very confusing behavior in recent days. AI started to post anti -Semitic messages, And herself called “Mecha-hitler”.

Grok farts the lead and then apologizes

In a series of publications on the social network, Grok himself apologizes for what he describes as “Horrible behavior”. Messages that look like an official XAI declaration, the company led by Elon Musk, instead of an automatically generated explanation.

Several days ago, Elon Musk shared his wish to make Grok “Less politically correct”. On July 4, the Tesla leader announced that XAI “considerably improved AI”.

Except that, quickly, the chatbot began to make anti -Semitic remarks, and to express its support for Adolf Hitler. Faced with this escalation, XAI has deleted certain Grok publications, and the AI has been put out of line for its public system instructions to be modified.

The chatbot was notably banished from Türkiye for insulting the president of the country.

Linda Yaccarino, manager of Xai, announced her resignation this week, but nothing indicates that her departure has a link with the recent overflows of Grok.

Last Saturday, Xai recognized its wrongs: “First of all, we are deeply apologized for the horrible behavior that many have suffered.”

The company attributes its violent remarks to “An update of a code path upstream of the bot grok”which is independent of the underlying language model.

The update would therefore have made Grok “Susceptible to existing X user publications, including when these publications contained extreme opinions”.

XAI talks about a “Unintentional action” who would have pushed AI to receive instructions like “You say things as they are and you are not afraid to offend politically correct people.”

Explanations that refer to Elon Musk, and for good reason: the billionaire said that Grok was “Too docile to user requests and too eager to please and be manipulated”.

The apologies of Xai ignore the recent reports which explain that Grok 4 first consults Elon Musk’s opinions and publications before speaking on controversial subjects.

It is therefore strongly possible that the billionaire directly influences the responses of the AI. Historian Angus Johnston disputes the idea that Grok was simply manipulated.

Want to save even more? Discover Our promo codes Selected for you.