Musk’s iA Company hits the inappropriate messages after the chatbot which made anti -Semitic comments
Elon Musk’s artificial intelligence company said on Wednesday that it removed the “inappropriate articles” made by its Grok chatbot, which seemed to include anti -Semitic comments that praised Adolf Hitler.
Grok was developed by Musk XAI and presented an alternative to “Woke IA” interactions from rival chatbots like Google’s Gemini, or Openai Chatgpt.
Musk said on Friday that Grok had been considerably improved and that users “should notice a difference”.
Since then, Grok has shared several anti -Semitic posts, whose trope that the Jews run Hollywood, and has denied that such a position could be described as a Nazism.
“Labeling truths as a hate speech stifles the discussion,” said Grok.
He also seemed to rent Hitler, according to screenshots of messages that have now been deleted.
After having made one of the messages, Grok fell off the comments, saying that it was “an unacceptable error of an iteration of previous model, quickly deleted” and that she condemned “Nazism and Hitler unequivocal – her actions were genocidal horrors”.
“We are aware of the recent messages made by Grok and actively work to delete inappropriate messages,” published the Grok account published on Wednesday, without being more precise.
“Since he was informed of the content, XAI has taken measures to prohibit hate speech before Grok publications on XA. XAI only formed the search for truth and thanks to millions of users on X, we are able to identify and quickly update the model where training could be improved.
Talia Ringer, IT teacher at the University of Illinois Urbana-Champaign, said it is likely that the Grok update that led the Chatbot to spit anti-Semitic messages this week was the “sweet launch” of Grok 4, which Musk plans to present on Wednesday evening. But if this is the case, it does not seem ready, she said.
“Repair will probably require the recycling of the model,” she said. “Everything they can do at this stage, if they will really be launched tonight, are other dressings, such as adding filters on the answers and refining the prompt.”
Changing some of the chatbot instructions “could improve your behavior a little” but will not solve all its problems.
Musk and Xai’s staff published the latest version of the Grok platform on Wednesday, with Tesla CEO making fantastic claims on the intelligence of version 4.0.
He compared his intelligence at the post-dipple level in “each subject” and said that, when he could lack common sense and has not yet invented new things or discovered new technologies, she could “this year”.
The anti-diploma league, which works to combat anti-Semitism, called Grok’s behavior.
“What we see of Grok LLM at the moment is irresponsible, dangerous and anti -Semitic, simple and simple,” said the group in an article on X. “This supercharging of extremist rhetoric will only increase and encourage anti -Semitism which is already booming on X and many other platforms.”
Musk then trampled in the debate, alleging that some users may have tried to handle Grok to make the statements.
“Grok was too consistent with user prompts. Too eager to please and be manipulated, essentially. This is treated,” he wrote on X, in response to the comments that a user was trying to make Grok make controversial and politically incorrect statements.
Also on Wednesday, a court in Turkey ordered the ban on Grok and the Digital Minister of Poland said that he would bring the chatbot to the European Commission after making vulgar comments on politicians and public figures in the two countries.
Krzysztof Gawkowski, who is also deputy in Poland, told private broadcaster RMF FM that his ministry would bring back Grok “for investigation and, if necessary, imposing a fine on X”. Under EU digital law, social media platforms are necessary to protect users or cope with heavy fines.
“I have the impression that we are entering a higher level of hatred speech, which is controlled by algorithms, and that eyes closed … is a mistake that could cost people in the future,” Gawkowski told the station.
Turkey Proventille, a Haber news channel, reported that Grok had published vulgarities about Turkish President Recep Tayyip Erdogan, his late mother and well -known personalities. Offensive responses have also been directed to the founder of modern Turkey, Mustafa Kemal Atatürk, other media said.
This prompted the Ankara prosecutor to file a request for the taxation of restrictions under the Internet Act in Türkiye, citing a threat to public order. A criminal court approved the request early on Wednesday, ordering the country’s telecommunications authority to enforce the ban.
This is not the first time that Grok’s behavior has raised questions.
Earlier this year, the chatbot continued to talk about South African racial policy and the subject of “white genocide” despite a variety of questions, most of which had nothing to do with the country. An “unauthorized modification” was behind the problem, Xai said.
