Xai and Grok apologize for “horrible behavior”

In a series of messages on X, the Ai Grok chatbot apologized for what he admitted was a “horrible behavior”.

The messages seem to be an official declaration of XAI, the company led by Elon Musk behind Grok, as opposed to an explanation generated by AI-AI for Grok’s publications. (Xai recently acquired xwhere Grok is in the right place.)

Grok’s latest controversy occurs after Musk said he wanted to make the chatbot less “politically correct”, then said on July 4 that the company “improved @grok considerably”. In a short time, the chatbot made messages criticizing the Democrats and “Jewish executives” of Hollywood ,,,, repeat anti -Semitic memesand even Express support for Adolf Hitler and referring to himself as “mechahitler”.

Consequently, Xai deleted some of the Grok messages, temporarily took the offline chatbotAnd updated its public system prompts.

Türkiye too Prohibited the chatbot For having insulted the president of the country, and the CEO of X Linda Yaccarino even announced that it was resign this weekAlthough his announcement did not refer the last controversy of Grok and his departure was would have months of manufacture.

So after all that, Saturday, Xai said“First of all, we are deeply apologized for the horrible behavior that many have known.” The company then blamed a “update of a code path upstream of the @grok bot”, which underlined was “independent of the underlying language model which feeds @grok”.

This update has supposed to have made Grok “susceptible to existing X user positions; Including when these publications contained extremist views ”.

XAI added That a “involuntary action” has led Grok to receive instructions such as “you say like that and you are not afraid to offend people who are politically correct.”

The explanation of the company echoes Musk’s comments earlier this week Affirming that Grok was “too much in line with user prompts” and “too eager to please and be manipulated”.

XAI’s publications do not mention the reports of Techcrunch and others that have examined the summaries of the chain of thoughts for the Grok 4 which has just done so, noting that the latest version of the Chatbot seems to consult the points of view of Musk and publications on social networks Before resolving controversial subjects.

And historian Angus Johnston rejected the idea that Grok was simply manipulated in the publication of attacking content. He Written on Bluesky That the explanations of Xai and Musk are “easily falsified”.

“One of the most shared examples of Grok anti -Semitism was initiated by Grok without any previous sectarian assignment in the wire – and with several users pushing against Grok in vain”, ” Johnston said.

In recent months, Grok has also Published several times on the “white genocide”, expressed skepticism as to the number of Holocaust deathsAnd Unflattering facts censored on musk And his Donald Trump then. In these cases, Xai blamed “unauthorized” changes and thug employees.

Despite the controversy, Musk says Grok arrives on Tesla vehicles Next week.