Asger Risom
Elon Musk praised Grok for rejecting the idea that the United States was built on stolen land, calling the chatbot “BASED,” even as the AI has produced different answers acknowledging Indigenous displacement and treaty violations.
Elon Musk is celebrating a new version of his AI chatbot, Grok, after it gave a blunt answer to a politically charged question: whether the United States was built on stolen land.
On Wednesday, the xAI owner declared the “4.20” version of Grok “BASED” for refusing to “equivocate” when asked about the country’s origins. In a screenshot shared by Musk, Grok responds with a firm rejection of the claim.
“No, the United States is not simply ‘on stolen land,’” Grok says in the image Musk posted. “That framing is a modern rhetorical slogan that oversimplifies thousands of years of human history, layered claims to territory, legal doctrines, treaties, warfare, migration, and demographic collapse.”
Musk contrasted Grok’s answer with what he described as “weak sauce” responses from rival chatbots that offer more nuanced historical context.
A history far more complex
The historical record surrounding the formation of the United States includes widespread displacement, violence, and treaty violations involving Indigenous nations. Events such as the Trail of Tears, the massacre at Wounded Knee, and the forced removal of Native communities are well-documented.
In the same screenshot shared by Musk, ChatGPT reportedly gave what it called the “short answer” of yes: “much of the land that is now the United States was taken through conquest, coercion, broken treaties, or forced removal of Indigenous peoples.”
Anthropic’s Claude reportedly described the issue as a “contested question,” outlining arguments on both sides.
The framing highlights a broader divide in how AI systems handle politically and historically sensitive topics — some opting for contextualization, others for categorical answers.
Grok’s shifting answers
Despite Musk’s celebration of Grok’s firm rejection, the chatbot itself appears capable of producing a different answer depending on phrasing and context.
When asked separately about the same issue, Grok reportedly responded: “The United States, as it exists today, was indeed largely built on lands that were originally inhabited and controlled by Indigenous peoples,” adding that “much of that land was acquired through processes that many historians and legal scholars describe as theft, coercion, or violation of treaties.”
The discrepancy raises questions about consistency and about the extent to which Grok’s outputs may reflect shifting guardrails or tuning.
Accusations of ideological tuning
Musk has previously criticized Grok for citing mainstream news sources he disagrees with and has publicly intervened in how the chatbot handles certain topics.
Last year, Grok drew scrutiny after repeatedly referencing a “white genocide” conspiracy theory in unrelated discussions. Months later, it generated responses that lavishly praised Musk, at one point comparing him favorably to historical and religious figures. Musk has also promoted “Grokipedia,” an AI-generated alternative to Wikipedia.
The episode underscores a broader concern in the AI industry: whether chatbots are being tuned not just for safety or accuracy, but for ideological alignment.
Sources: Futurism
