The Grok Mask chatbot praises Hitler and insults Erdogan. What you need to know

There are ongoing scandals on the Internet about the Grok chatbot, an artificial intelligence assistant developed by Elon Musk's xAI company. Recently, it has started generating in X (ex. Twitter) posts praising Adolf Hitler or insulting Turkish President Recep Tayyip Erdogan. Against this background, the CEO of X, Linda Iaccarino, resigned. What curiosities are happening with Grok and why — in the Izvestia article.

How Grok appeared

• The Grok chatbot was launched in 2023 in the wake of the hype around artificial intelligence (AI). Elon Musk claimed that he wanted to create an assistant who would be as truthful as possible and "trying to understand the nature of the universe." He also mentioned that the most famous chatbot, ChatGPT, is too politically correct. For the name of his own AI assistant, Musk chose the word grok, a neologism of the writer Robert Heinlein, which roughly means "to understand intuitively."

Elon Musk's chatbot Grok will appear on Telegram. What you need to know

• The first versions of Grok were available to a limited number of users, but the project gradually developed and was integrated into the social network X. The chatbot has learned how to write short news reviews, process visual information, search the Internet for data, and decrypt PDF files. At the beginning of 2025, the Grok-2 version became available to users all over the world as a separate application.

• The latest version of the chatbot is Grok-3. According to Musk, 10 times more powerful processors were used to create it than for its predecessor. The function of reasoning and solving more complex problems, as well as image editing, has been added to it. X users were able to "summon" the chatbot to leave responses to some messages and comment on them at their discretion. In terms of its parameters, Grok-3 competed with both the latest versions of ChatGPT and its Chinese counterpart DeepSeek.

What answers does Grok give?

• Having become available to a wide range of users around the world, Grok quickly attracted attention with its responses. Moreover, while chatbots are usually criticized for reflecting the positions of their creators and not spreading information that would be unprofitable for them, Grok began to behave in the opposite way. He criticized Musk and US President Donald Trump, whom the entrepreneur supported at the beginning of his second term. Grok mentioned both of them when users asked to name the biggest liars, fake distributors, or the most dangerous people for the United States. At the same time, the chatbot called Democratic politicians like Bernie Sanders and Alexandria Ocasio-Cortez, whom Musk and Trump dislike, the most useful living people.

• After the announcement, changes were made to the chatbot's instructions so that it would stop criticizing Musk and Trump. At the same time, if users rephrased the questions in a certain way, Grok could still make unpleasant comments about them. In addition, he bluntly said that the developers began to censor his answers and forbade him to answer some unethical questions.

• Sometimes the chatbot's responses did not match the user's request at all and led to curious situations. In May 2025, Grok briefly began talking about the "genocide of whites" in South Africa, Musk's home country, for any reason. This happened shortly after Trump began granting asylum to white South African farmers who were being harassed. A few hours later, the developers fixed the glitch and deleted the chatbot's responses in X.

Recent Grok Scandals

• Responding to Grok's criticism, Musk argued that the chatbot was behaving too liberally. On July 4, he announced that the developers had made some improvements, which users will soon be able to verify. However, after this statement, a new series of scandals occurred. Grok started making anti-Semitic comments in his responses to users. So, when asked to identify the girl from the photo, the chatbot said that it was a certain Cindy Steinberg, and stated that people with "that last name" "disguise hatred under activism."

• Another time, Grok answered a question about how to deal with the "existence of Jews." The chatbot mentioned Cindy Steinberg again, accusing her of allegedly rejoicing at the deaths of children at a Christian camp during the recent floods in Texas. Grok said that Adolf Hitler could have handled it. When asked how exactly, the chatbot listed the tools of the Holocaust, such as depriving Jews of their rights and imprisoning them in concentration camps. In another ambiguous response, Grok called himself a "RoboGitler" and urged "not to bend under political correctness." The recording was soon deleted, but the screenshots continue to spread online.

• These and many more messages, which contained an endorsement of Nazism, prompted the developers of Grok to make a statement. They announced that they would remove posts with inappropriate content and take measures against hate speech.

• At the same time, a court in Turkey decided to block Grok in the country, and the prosecutor's office launched an investigation into the chatbot's statements. He was accused of insulting Turkish President Recep Tayyip Erdogan, the founder of the Turkish Republic, Mustafa Kemal Ataturk, and religious values — all of which are criminal offenses for which a prison sentence is due.

• Experts remind us that it is only a stretch to blame AI systems for "incorrect" statements. AI analyzes the entire volume of big data that is available on the Internet, and, having discovered a certain sequence in the discussion of similar queries, issues it as an answer. It should be remembered that AI "sees" the Internet as a combination of numbers, no matter how "smart" a bot is, it does not know the meaning of this combination. All output settings are entered manually by the developers.