Though temporarily overtaken by events, the conflict between Anthropic and the Department of War, which broke into open lawfare in February, has made the problem of sovereignty in the age of AI unmistakable. Tensions began with Trump Administration concerns that the post-Rawlsian ethical preoccupations of Claude’s creators made the model’s worldview essentially woke, a reversal of Biden-era obsessions with algorithmic bias. Can Effective Altruists, apparently committed to the equal consideration of interests across all space-time and species, actually be patriots, who prioritize the ties of particular places and peoples?
Things fell apart, reportedly, after representatives of Anthropic expressed worry that Claude’s application in classified settings as part of Palantir’s military systems had perhaps violated their internal use-case red lines. Secretary of War Pete Hegseth took the position that, while no such lines had been crossed, a tool for American military use can have no red lines except those determined by American law. Words flew hot, then orders, then lawyers.
A relatively limited supply-chain risk designation has ping-ponged through the courts. Most recently a California U.S. district judge gave a point in Anthropic's favor. After an expected appeal, the technical legal considerations will be eventually determinative, but to most observers paying attention they are largely beside the point. Both the heat and light of this flare up came from that fundamental question: Who decides the exception? In a self-governing republic, and dealing with technologies that are described by their own creators as dual-use in ways comparable to splitting the atom, can unelected businessmen really be allowed to control frontier AI’s employment and ends?
Whether AI products are scaling into the age-ending equivalent of a nuclear weapon is speculation beyond the scope of a brief theoretical essay like this. The question of how to govern this technology, and what its significance is to the American project, is one worth grappling with on present terms, in light of current technology. At heart, the human hope for AI is responsible mastery—accountable use of powerful, trustworthy tools for our benefit. The alternative, as we are constantly reminded by boosters and doomers alike, is slavery in a panopticon perfected. The doubting public will believe in the hopeful vision when they see it, which means the state, as their representatives, must take the lead in adopting and mastering these technologies to fulfill its best ends.
If the U.S. Federal Government does not master AI for the renewal of the Constitutional system then it resigns the country to AI technology, politically speaking, exacerbating rather than ameliorating the problem of rule by the few in the American regime. This would metastasize almost certainly to the point of civil stasis in both classical senses of the word, violent conflict and social sclerosis, a dying order.
Imagine a DOGE, but different. A DOGE that actually was capable of being a real U.S. Digital Service, and a DOGE that was worthy of the name Department of Government Efficiency. No longer would the president’s appointees be forced to cling to a golden thread in the maze of the bureaucracy, eking out a spending cut here and there or bleeding for a bulletproof reduction in force. Rather, the ontology of the administrative apparatus would be clarified; there would be rectification of names. From the West Wing and the Eisenhower Executive Office Building all the president’s men would man immersive dashboards like so many mecha pilots. The Resolute Desk would fill the Oval Office with a holographic command and control display that bound the many-armed beast of the state into a double-bridled behemoth, gentle as an ox. Bypassed by an efficient, responsive, legible, auditable system of agents immediately subject to politically accountable principals, bureaucratic bloat could be left to rot on the vine, pruned at the convenience of the duly elected. We the people’s representative executive would execute directly; elections would mean something.
How the government should regulate AI and AI companies is a distinct, if related, question to how the government should use AI technology, and in hope of human mastery the latter must be answered first. Hence this tract. The meta narrative of progress and reaction that defines so much of our cultural detritus hinders rather than helps clarifying the task of statesmen in our technological moment. It elides the more fundamental consideration of bad and good. What does it mean to govern well, in this technological moment? What poorly? Moreover, the blinders placed on us by our fixation on progress prevents us from taking our lights from even the recent past. The focus on rupture rather than continuity from humanists and AI CEOs alike immobilizes instead of informing a coherent response to the needs of the nation.
If we set aside spectacle and speculation, we can recognize that even the most psychotically charged possibilities of this period of technological development are emergent from a historical continuity, an acceleration of trends that have defined the last century. Concerns about technicity, corporate power, and miseducation suffocating meaningful community, culture, and self-government are all central fears of our post-modern era—the civilizational counterpoise to material abundance. AI anxiety gives these a finer point and closer proximity for the average literate American, but these feelings are working out a now well-established tradition in our national life, and when framed as such come accompanied by traditional responses.
One major throughline in our civic story, a perennial problem of political philosophy and matrix for understanding political power, appears to have arrived at a moment of crisis simultaneously to information technology’s virtual apotheosis: namely the regime relationship between the one, the few, and the many. The American founders, as modern as they were, recognized that the classical conflict between these castes of public life persisted even in the midst of societies transformed by new scientific knowledge and technical breakthroughs. They constructed a mixed regime, a republic, in an effort to harness these ur-factions together to form a more perfect union promoting the general welfare. The decisive feature of the Federalist’s Constitutional coup against the failed Articles of Confederation was, following the lessons of Great Britain’s civil wars and upheavals, the alliance of a monarchic chief executive with the national populace, in counterweight to a pre-17th Amendment Senate representing the sectional interests of state elites. The one and the many would check the few; a legalist king and mercantile commons keeping the landed lords in check, and vice versa, in the name of liberty.
As much as attempting an extended republic at all, it was the uniting of authority, responsibility, and power in the elected presidency as both head of state and head of government that marked the American experiment as a new political science. The original techno-industrial nationalist Alexander Hamilton wrote in Federalist no. 70, “Energy in the executive is a leading character in the definition of good government. It is essential to the protection of the community against foreign attacks; it is not less essential to the steady administration of the laws; to the protection of property against those irregular and high-handed combinations which sometimes interrupt the ordinary course of justice; to the security of liberty against the enterprises and assaults of ambition, of faction, and of anarchy.” From Jefferson to Jackson, from Lincoln to Roosevelt, from Nixon up to the present day, the foundings and refoundings of the American republic have been chiefly driven by executive responses to new conditions of industry, resources, finance, and labor created by the developments of technology.
The renewed interest in developing and implementing the unitary executive reading of Article Two in the last few decades, as the emotions of Watergate subside, spring from the recognized failure of managerialism. Depoliticization and technocratic control have not eliminated the irresponsible rule of the few or checked oligarchic power as it shifted from ownership of land to capital to data, but only replaced a gilded unaccountable elite with a regime of beige public-private partnership. The few still rule, but from gray-carpeted offices, erecting no monuments but a world more leveled and fungible. Gold-lover that he is, President Trump has been elected and reelected in an appeal to political accountability, the middle finger to this alienating establishment. Nationalism, and the many’s grasping for the one as their champion, has always been a self-conscious people’s recourse against the threat of homogenization, whether in response to visible empires, the force of global markets, or the growing sense that an emergent System seeks to absorb and rationalize the totality of human life.
The need Progressives sought to meet with the manufacture of an administrative elite managing structures of scale and complexity beyond normal human capacities persists, however. Technicity has accelerated, not abated. The few still undermine the mixed regime that gives the many a chance of self-government. The populist moment continues then, as well, unresolved, in a quest for political agency. AI technology presents, along with its challenges, a new tool for completing the attempted Constitutional renewal, and a protective mediation between the American people and the forces they face. In a spirit of seriously but not literally, though perhaps more literally with increased attention, an analogy can be drawn between building and using AI infrastructure that is open source, localized, and personalized vis a vis the SAAS, hyperscaled, and proprietary subscription model and the renewal of a Constitutional government vis a vis the Administrative State’s absorption of Executive and Legislative function abetted by the Judiciary.
As a culmination of a century of developments in information technology, contemporary AI presents a made to measure tool for restoring the responsive executive, who may by injection of sufficient energy into American life elicit a pulse from the other branches of Federal Government and revivify the Constitutional system. Only then, and through mastery of the tools of the technological elite, can we hope for a representative republic capable of checking oligarchic power and mediating between the American citizenry and the upheavals of this brave new world. The response to regulatory capture and special privileges in our algorithmic age can be the regulatory clarity, legal consistency, and digital transparency of machine intelligence under human direction.
A signer of the Declaration, participant in the Constitutional Convention, and original Supreme Court associate justice, James Wilson said at the ratifying convention for Pennsylvania: “The executive authority is one.
We may discover from history, from reason, and from experience, the security which this furnishes. The executive power is better to be trusted when it has no screen. Sir, we have a responsibility in the person of our President; he cannot act improperly, and hide either his negligence or inattention; he cannot roll upon any other person the weight of his criminality; no appointment can take place without his nomination; and he is responsible for every nomination he makes. We secure vigor. We well know what numerous executives are. We know there is neither vigor, decision, nor responsibility, in them.”
Today, as you read this on a smartphone or computer, we may amend this statement slightly, and say, “The executive power is better to be trusted when it has many screens indeed.”