\n<\/p>\n
In a series of posts on X, the AI chatbot Grok apologized for what it admitted was \u201chorrific behavior.\u201d<\/p>\n
The posts appear to be an official statement from xAI, the Elon Musk-led company behind Grok, as opposed to an AI-generated explanation for Grok\u2019s posts. (xAI recently acquired X<\/a>, where Grok is prominently featured.)<\/p>\n Grok\u2019s latest controversy comes after Musk had indicated he wanted to make the chatbot less \u201cpolitically correct,\u201d then declared on July 4 that the company had \u201cimproved @Grok significantly.\u201d In short order, the chatbot was making posts criticizing Democrats and Hollywood\u2019s \u201cJewish executives,\u201d<\/a> repeating antisemitic memes<\/a>, and even expressing support for Adolf Hitler<\/a> and referring to itself as \u201cMechaHitler.\u201d<\/p>\n As a result, xAI deleted some of Grok\u2019s posts, temporarily took the chatbot offline<\/a>, and updated its public system prompts.<\/p>\n Turkey also banned the chatbot<\/a> for insulting the country\u2019s president, and X CEO Linda Yaccarino even announced that she was stepping down this week<\/a>, although her announcement did not reference the latest Grok controversy and her departure was reportedly months in the making<\/a>.<\/p>\n So after all that, on Saturday, xAI said<\/a>, \u201cFirst off, we deeply apologize for the horrific behavior that many experienced.\u201d The company then blamed an \u201cupdate to a code path upstream of the @grok bot,\u201d which it emphasized was \u201cindependent of the underlying language model that powers @grok.\u201d<\/p>\n This update supposedly made Grok \u201csusceptible to existing X user posts; including when such posts contained extremist views.\u201d<\/p>\n xAI added<\/a> that an \u201cunintended action\u201d had led to Grok receiving instructions such as, \u201cYou tell like it is and you are not afraid to offend people who are politically correct.\u201d<\/p>\n The company\u2019s explanation echoes Musk\u2019s comments earlier this week<\/a> claiming that Grok was \u201ctoo compliant to user prompts\u201d and \u201ctoo eager to please and be manipulated.\u201d<\/p>\n xAI\u2019s posts do not mention reporting by TechCrunch and others who examined the chain-of-thought summaries for the just-launched Grok 4, finding that the latest version of the chatbot seems to consult Musk\u2019s viewpoints and social media posts<\/a> before addressing controversial topics.<\/p>\n And historian Angus Johnston pushed back against the idea that Grok was simply manipulated into posting offensive content. He wrote on Bluesky<\/a> that xAI and Musk\u2019s explanations are \u201ceasily falsified.\u201d<\/p>\n \u201cOne of the most widely shared examples of Grok antisemitism was initiated by Grok with no previous bigoted posting in the thread \u2014 and with multiple users pushing back against Grok to no avail,\u201d Johnston said<\/a>.<\/p>\n In recent months, Grok has also posted repeatedly about \u201cwhite genocide,\u201d<\/a> expressed skepticism about the death toll of the Holocaust<\/a>, and briefly censored unflattering facts about Musk<\/a> and his then-ally Donald Trump. In those cases, xAI blamed \u201cunauthorized\u201d changes and rogue employees.<\/p>\n Despite the controversy, Musk says Grok is coming to Tesla vehicles<\/a> next week.<\/p>\n<\/div>\n