Trump and Musk both want to control America's history

Trump Orders Smithsonian Signage on Exhibit Accuracy; Musk Announces AI-Generated Odyssey Project

Donald Trump issued an executive order on July 24 directing the placement of exterior signage at Smithsonian Institution museums. The signs would notify visitors of what the order describes as “inaccurate information” in current displays and direct them to other resources for information on American history. Two days earlier, Elon Musk posted on X that the generative AI model Grok would produce a feature-length film adaptation of Homer’s “The Odyssey” by the end of the year, described as “historically accurate and true to the art of Homer.”

The two announcements occurred within the same week but addressed separate domains: one involving federal museums and the other a commercial AI product. The Smithsonian Institution operates under a board of regents that includes the chief justice of the Supreme Court, the vice president, and members of Congress. Its museums receive federal appropriations but maintain independent governance structures established by statute. Legal analysts have noted that executive orders directed at the Smithsonian face practical and legal constraints because the institution is not an executive branch agency.

Musk’s announcement concerns the capabilities of the Grok model, which is developed by xAI. No technical details were provided on how the model would ensure historical accuracy for an ancient epic poem whose surviving texts date to the eighth century BCE and whose events are set in an earlier period. AI image and video generation systems operate by identifying statistical patterns in large training datasets. These systems can reproduce common visual conventions found in existing sources but have documented limitations when applied to periods with sparse or contested visual records.

On July 21, historian David Blight of Yale University testified before the House Oversight Subcommittee on the Smithsonian. He emphasized the value of primary-source research and the role of interpretive debate in historical scholarship. Blight referenced concerns about centralized control over historical narratives, citing author Toni Morrison’s comments on “statist history” and “sanctioned ignorance.”

The same week, Michael Kratsios, the White House science adviser, outlined a proposal to shift a larger share of federal science funding toward private corporations and philanthropic organizations. Kratsios previously worked at an investment fund associated with Peter Thiel and at the Department of Government Efficiency. The proposal lists AI, quantum computing, and robotics among priority areas. On the same day, Kratsios announced grant recipients under the Genesis Mission, an administration AI initiative.

Federal support for the humanities has historically been smaller than support for the physical sciences. Museums and archives have entered into partnerships with technology companies for digitization projects and visitor-experience tools. Examples include fellowships supported by the Schmidt Futures foundation at the Metropolitan Museum of Art and the Musk Foundation’s backing of the Scroll Prize, which applies machine-learning techniques to the reconstruction of carbonized papyri from Herculaneum. Some AI firms have also acquired and scanned physical volumes for training data, a practice that has drawn criticism from preservationists.

Training datasets for current generative models are drawn from digitized collections whose composition reflects the acquisition priorities of past institutions. Historians have long documented that museum and archive holdings over-represent certain regions, languages, and social groups while under-representing others. When models are trained on these collections, outputs tend to reproduce the distribution of the source material rather than correct for documented gaps.

The Smithsonian signage directive and the Grok film announcement remain separate initiatives. One involves an executive directive whose implementation depends on cooperation from an independently governed institution. The other involves a commercial claim about future AI output whose technical feasibility has not yet been demonstrated. Both occur against a backdrop of ongoing discussion about the sources of funding for research and public cultural institutions and the properties of large-scale AI systems when applied to historical subjects.