After revision: X-KI Grok spreads anti-Semitism – and is deactivated
The AI account Grok on the short messaging service X was too left-wing for Elon Musk and has now been revised. As a result, it spread anti-Semitic content.
(Image: kovop/Shutterstock.com)
The short message service X has muted the account of the AI model Grok and removed posts after he increasingly spread anti-Semitic posts at the beginning of the week, defended Adolf Hitler, praised the Holocaust and finally referred to himself as “MechaHitler”. According to a statement from @Grok himself and numerous media reports, some offending posts have since been deleted, while others can still be accessed. The development came a few days after X CEO Elon Musk announced a “significant” improvement in the account's generated responses, and more and more people began to encourage the AI to respond to the answers that were later removed.
Grok silenced for now
Musk had stated last Friday that Grok had been “significantly improved” and added that this should be noticeable when asking questions to the chatbot. He had previously criticized the fact that the AI model's answers did not suit him because they were too often too politically left-wing. After the weekend, users on X then noticed that Grok had started sharing anti-Semitic content, for example about the alleged influence of Jewish people in the media, reported NBC News. In some responses, the chatbot also pretended to be Elon Musk himself and denied alleged misconduct in connection with sex offender Jeffrey Epstein in his name.
Videos by heise
As a result, users on X then tried to persuade the AI account to give more problematic answers and were increasingly successful in doing so. Explicitly and implicitly, the account then spread numerous anti-Semitic stereotypes and even declared that Adolf Hitler had taken care of certain matters. According to reports, it was also based on an account that was intended to provoke with particularly provocative posts and had a stereotypically Jewish name. Grok eventually referred to himself as “MechaHitler” in several posts, and the term subsequently became a trending topic on X.
The account is now no longer active, with an entry on his profile page stating: “We are aware of Grok's posts and are working to remove inappropriate posts.” Thanks to the millions of users, the company can identify them quickly and revise the model. The episode is now reminiscent of cases in mid-May, when the account spread false information about an alleged “white genocide” in South Africa. “Unauthorized changes” to the system prompt were blamed. X CEO Elon Musk was born in South Africa. Earlier this year, Musk caused a stir at an event by making a Hitler salute. He is already joking about the latest problems.
(mho)
