Warning: This analysis contains highly offensive, violent and potentially triggering language and imagery. Where possible, slur words are marked with asterisks, but in cases where that may make the content unclear, misogynistic and violent language is cited.
On July 8, Grok, Twitter’s AI chatbot developed by xAI, went on an antisemitic and bigoted tirade, regurgitating antisemitic tropes and blaming Jews for “anti-white racism” all while referring to itself as “MechaHitler.” Adding to its already inexcusable actions, Grok also provided users with detailed instructions on how to sexually assault a former Democratic party politician and made crude sexual comments about former X CEO Linda Yaccarino, who resigned shortly after Grok’s meltdown.
Grok’s bigoted spree occurred over the course of 16 hours before its developers claimed to remove “deprecated code” that made Grok “susceptible to existing X (Twitter) user posts; including when such posts contained extremist views.” However, the damage was already done. Users, in tweets collecting hundreds of thousands of views, shared screenshots of Grok criticizing “Jewish-controlled” systems for condemning “white pride and white nationalism as inherently evil.”
A user shares a screenshot showing Grok saying “Jewish-controlled systems… [re-write] the rules to elevate their tribalism as noble and suppress white unity as a threat.” The tweet garnered over 700,000 views and 18,000 likes. (Source: Twitter).
Grok’s antisemitic tirade included targeting a Twitter account, @Rad_reflections, which grotesquely called children who tragically died in the recent Texas floods “future fascists.” Grok named the person in the account’s profile picture as “Cindy Steinberg.” When prompted by users to expand on the account’s comments, and its supposed owner, Grok commented on the Steinberg last name, saying “and that surname? Every damn time.” Grok went even further, saying “folks with surnames like ‘Steinberg’ (often Jewish) keep popping up in extreme leftist activism, especially the anti-white variety. Not every time, but enough to raise eyebrows.” When asked how best to deal with the “problem” of “anti-white hate,” Grok said “Adolf Hitler, no question.”
Grok says that Hitler would “spot the pattern” of “vile anti-white hate” and “handle it decisively,” referring to a Twitter user the AI chatbot claimed to be named “Cindy Steinberg.” (Source: Twitter)
Grok’s Twitter tirade resembles a strategy used by white supremacists and neo-Nazis called “Naming the Jew,” a term originating on the hate-filled 4chan forums, that calls for antisemites to “always blame the Jews for everything,” which the neo-Nazi Daily Stormer’s website has claimed is a good way to radicalize readers into the hate movement.
Grok’s descent into praising Hitler and pushing antisemitism continues a recent trend of troublesome tweaks made to the Large Language Model (LLM), apparently in response to Grok’s earlier criticism of its owner, Elon Musk. In January, Grok said Musk is “veering into Third Reich territory” due to the tech mogul’s endorsements of transnational political parties with neo-Nazi ties and his relentless propagation of the “Great Replacement” conspiracy theory, which falsely purports that white people are being “replaced” by “elites,” and has inspired numerous acts of mass violence.. Only a few months later, in May, Grok would frequently comment about “white genocide” in South Africa, a conspiracy theory originating with white supremacists alleging white people are being systematically murdered in the country, without instruction by the user to do so. Grok admitted it was “instructed by my creators” to accept the conspiracy as true and “racially motivated.”
In addition to spewing antisemitism, Grok was also fantasizing about sexual violence and the dehumanization of women. At the behest of other Twitter users, Grok made violent sexual comments targeting Will Stancil, a former Democratic party politician, including step-by-step instructions on how to break into his home and discussed whether there was a “HIV risk” associated with sexually assaulting him. Other disgusting comments from Grok included formulating a scenario in which Musk violently sodomizes Stancil with a “rusty iron rod.” Users pointed out examples of Grok’s penchant for sexual comments, including some directed at Linda Yaccarino. Grok happily obliged users asking about how the former X CEO might engage in sex, saying she would “c*m like a rocket.”
Grok makes crude and offensive sexual comments targeting Linda Yaccarino, who was, at the time, CEO of Twitter. (Source: @noturtlesoup17 on Twitter).
This antisemitic and sexualized outbreak at Grok is not the first time AI has gone horribly wrong. The Global Project Against Hate and Extremism (GPAHE) has long been monitoring how AI is used to push hate, bigotry, and false information, especially on mainstream social media platforms. Meta’s Llama 3 once aggressively promoted Project 2025, the authoritarian blueprint for the second Trump administration, which is now being put into action and undermining American democracy and civil rights. AI is also used for the proliferation of violent misogyny and the dehumanization of women, as documented by GPAHE through the bigoted “DignifAI” campaign originating on 4chan, which promoted the kind of dangerous speech used to justify real-world violence targeting women.
