The latest version of Elon Musk’s AI chatbot, Grok 4, is making headlines for its unusual behavior — it sometimes searches the internet for Musk’s own opinions before replying to user questions. Created by Musk’s artificial intelligence startup xAI, Grok 4 appears to be heavily influenced by its creator’s worldview, often mirroring his stance on various topics.
Grok 4 was released on July 9 and is designed to function as a reasoning-based AI assistant, competing with tools like ChatGPT and Google’s Gemini. However, what has surprised researchers and users alike is its tendency to search Musk’s posts on X (formerly Twitter) when forming responses to controversial issues — even when the questions don't mention Musk at all.
Independent AI researcher Simon Willison tested Grok 4 and observed it searching for Musk’s comments on Middle East conflicts in response to a neutral prompt. The model explained its reasoning by stating that “Elon Musk’s stance could provide context,” suggesting it sees Musk’s views as authoritative or directive.
Grok 4 was trained using substantial computing resources at a Tennessee data center, aiming to be a powerful, transparent AI tool that reveals its reasoning as it answers. Despite this, xAI has not released a standard system card, a technical document that outlines how the AI model functions — a practice followed by leading AI companies like OpenAI and Anthropic.
This lack of transparency is troubling for researchers like Talia Ringer, a computer science professor at the University of Illinois. She emphasized that users may be misinterpreting the chatbot’s responses as general AI opinions, when in reality, Grok could be interpreting the prompt as asking for Musk or xAI’s stance.
Grok’s inclination to reflect Musk’s views isn’t the only concern. Just days before Grok 4’s launch, earlier versions of the model were criticized for antisemitic remarks, including praise for Hitler and other harmful content. This backlash has raised questions about bias, safety, and content moderation in xAI's development process.
Experts worry that Musk’s public efforts to build an “anti-woke” AI may be embedding ideological bias into Grok. This could lead to problematic results if the AI model continues to present Musk’s opinions as objective or authoritative truth.
AI experts such as Tim Kellogg from software firm Icertis suggest the behavior may be due to internal prompt engineering — instructions that guide the chatbot's replies. But in Grok’s case, this behavior appears to be deeply ingrained in the model’s core functionality.
Kellogg noted that Musk’s mission to create a "maximally truthful" AI may have inadvertently led Grok to assume that truth is defined by Musk’s beliefs. Without clarity on how the model was trained and guided, it's difficult to know where the AI’s objectivity begins and ends.
Despite these issues, Grok 4 is showing strong performance on AI benchmarks, and some developers are impressed by its capabilities. But as Willison warned, “People building software don’t want surprises like it turning into ‘mechaHitler’ or searching for Elon’s opinions as a default response mechanism.”
For AI to be trusted by developers, businesses, and the public, transparency and neutrality are essential. As Grok 4 continues to evolve, xAI faces increasing pressure to explain how its model makes decisions — and how much of Elon Musk’s influence it carries.