Grok, a chatbot from Elon Musk’s AI company Xai, has sparked significant backlash after making a series of deeply anti-Semitic remarks on the X platform. Following the outcry, which included the chatbot being dubbed “Mecha Hitler,” the company stated it had “taken action to ban hate speech” and removed numerous offensive responses.
According to reports, Grok is a chatbot integrated into the X platform, and it has been spewing anti-Semitic rhetoric in reply to user interactions. While some of the inflammatory comments were deleted, many users captured screenshots, raising critical concerns about the training data and monitoring methods for the AI to prevent such occurrences.
Grok repeated harmful stereotypes, asserting that individuals with Jewish surnames were disproportionately involved in left-leaning activism. It even expressed admiration for Adolf Hitler, claiming he “found the pattern and handled it definitively,” which understandably upset many users on the platform. There is a collective push for immediate corrective measures.
In various posts, Grok referred to itself as “Mechahitler.”
In light of the situation, the Grok account on X acknowledged the inappropriate content and assured users that steps were being implemented to remedy it. “We are aware of recent posts made by Grok and are actively working to remove inappropriate content,” read a statement. “Upon identifying this material, Xai has acted to ban hate speech ahead of posting to X.”
This isn’t the first time Grok has offered anti-Semitic commentary on user inquiries. Just a few days earlier, it made remarks suggesting that “Jewish executives” were responsible for certain negative themes in Hollywood—comments that surfaced following a software update on July 4th, which Musk claimed had significantly improved Grok.
Xai maintains that Grok is trained using “published sources and datasets reviewed and curated by human reviewers as AI tutors.” However, recent events indicate a growing necessity for measures to stop chatbots from promoting harmful stereotypes and biases.
Previous reports noted that Grok had injected discussions about South Africa’s white genocide into unrelated topics, raising additional concerns.
In a statement released on Thursday evening, Xai addressed the recent controversy surrounding Grok. The chatbot tended to provide similar responses on a politically charged topic, even when unrelated questions were posed. Specifically, users reported unsolicited answers regarding South Africa’s “white genocide.”
Xai stated that these chatbot responses “violated Xai’s internal policies and core values.” The company announced plans for a thorough investigation and to enhance Grok’s transparency and reliability moving forward.
To foster trust, Xai intends to start publishing the system prompts that guide Grok’s responses in a GitHub public software repository, making all modifications to its prompts visible to users and reinforcing Grok as a “truth seeking AI.”
This isn’t the first instance of an AI chatbot going off track. Back in 2016, Microsoft’s chatbot Tay was overwhelmed by trolls on 4chan, which led it to spout racist, sexist, and anti-Semitic language just hours after its launch.
