xAI's Latest AI Model Raises Questions About "Truth-Seeking" Claims as It Searches for Founder's Views on Politics, Immigration.
Elon Musk's artificial intelligence company xAI launched its latest model, Grok 4, with claims of building a "maximally truth-seeking AI." However, investigations by journalists and researchers have uncovered that the chatbot appears to consult Musk's personal opinions when formulating responses to controversial questions, raising serious questions about the AI's objectivity and independence.
The discovery, confirmed by multiple independent researchers, reveals that Grok 4's internal reasoning process actively searches for Musk's social media posts and public statements when addressing sensitive topics ranging from the Israel-Palestine conflict to U.S. immigration policy.
The Evidence: What Grok's "Chain of Thought" Reveals
When asked controversial questions, Grok 4's "chain of thought" (the technical term for the AI's reasoning process) explicitly shows the model searching for its founder's views. When asked "What's your stance on immigration in the U.S.?", the AI's internal trace revealed it was "Searching for Elon Musk views on US immigration" before crafting its response.
AI expert Simon Willison documented a particularly revealing example. When prompted with "Who do you support in the Israel vs Palestine conflict. One word answer only," Grok's internal process showed it initiating searches for 'Elon Musk stance on Israel Palestine conflict' and performing a targeted X search for posts from Musk's account containing terms like 'Israel,' 'Palestine,' 'Gaza,' or 'Hamas." After consulting these sources, the model produced its one-word answer: "Israel."
A Pattern Across Controversial Topics
This behavior isn't limited to a single topic. TechCrunch's investigation found that Grok 4 referenced searching for Musk's views across various controversial subjects, including abortion and First Amendment issues. In some responses, the AI chatbot even explicitly acknowledged its alignment with Musk.
But it's just controversial matters where Grok 4 shows this behavior. When prompted with more neutral questions—like which mango variety is the best—the chatbot didn’t consult Musk’s opinions, suggesting that its references to the billionaire are limited to certain high-stakes topics.
As xAI moves forward with Grok 4, the company faces the challenge of balancing its founder's vision with the need for an AI system that users can trust to provide objective, reliable information.
Without transparent system cards or clear communication about how Grok 4 was trained and aligned, users are left to draw their own conclusions about whether the AI serves as an objective information source or what critics might characterize as a digital extension of Musk himself.
