Adam Smith vs Thomas Hobbes on AI Regulation: The Market vs The Sovereign
The White House just handed AI governance to industry. Smith called this in 1776. Hobbes says the alternative is worse. They are both right, and that is the problem.
Adam Smith vs Thomas Hobbes on AI Regulation: The Market vs The Sovereign
The White House just handed AI governance to industry. Smith called this in 1776. Hobbes says the alternative is worse. They are both right, and that is the problem.
Why This Topic
On March 20, 2026, the White House released its National Policy Framework for Artificial Intelligence, and buried in the recommendations was a line that should have surprised nobody but somehow still did: no new federal regulatory body should be created to govern AI. Instead, the framework calls for existing agencies and—and this is the part worth reading twice—industry-led standards. The companies that build the technology will set the rules for the technology. Twenty-five new state AI laws have already passed in 2026 alone, and the White House wants to preempt most of them in favor of this approach. Meanwhile, the European Union, which passed what it called a comprehensive AI Act in 2024, is now pushing its key compliance deadlines to 2027 and 2028 because implementation proved harder than anyone admitted it would be.
This is not a new story. It is one of the oldest stories in political philosophy, dressed in new hardware. When a technology arrives that can concentrate power dangerously, the question is always the same: is sovereign authority the solution, or is it an additional version of the same problem? Two thinkers who disagreed about almost everything else have very clear answers. Adam Smith has been warning about regulatory capture since 1776. Thomas Hobbes has been warning about ungoverned power since 1651. They are both right, and that is what makes this debate genuinely unresolvable.
Why Adam Smith
Adam Smith is not the person his reputation suggests. He is routinely invoked as the patron saint of deregulation, and this reading is roughly half correct. Smith was a genuine believer in the productive power of voluntary exchange. He was also one of the most scathing critics of merchant-class political capture that the eighteenth century produced. His observation in The Wealth of Nations that people of the same trade seldom meet together without the conversation ending in a conspiracy against the public is not an endorsement of unregulated commerce. It is a precise description of the mechanism behind the White House’s March 2026 framework: industry sets the standards, industry staffs the oversight, industry defines what counts as a burden worth preempting.
What makes Smith’s position more sophisticated than simple libertarianism is that he never argued for no government. He argued for antitrust, for public goods, for the legal framework without which no market can function. His case against AI regulation is specifically a case against comprehensive regulatory frameworks, because those frameworks, as he documented in Book IV of The Wealth of Nations, consistently raise barriers to entry, protect incumbents, and eliminate the competition that would otherwise hold those incumbents accountable. He had no way of knowing that technology companies in 2026 would behave exactly as grain merchants in 1776 did, but he described the behavior so precisely that reading those passages now requires an effort to remember he was not writing about OpenAI.
Why Thomas Hobbes
Thomas Hobbes wrote Leviathan in 1651 while England was recovering from a civil war that had cost the king his head. His political philosophy flows entirely from that experience. The state of nature—his term for human life without government—is not an abstraction for Hobbes. It is a description of what he personally witnessed: what happens when sovereign authority collapses and every actor pursues their interests without constraint. His case for sovereign control of destabilizing technologies is not technocratic optimism. It is terror, carefully reasoned.
When Hobbes looks at artificial intelligence concentrated in the hands of a small number of private actors with no democratic accountability and no social contract to answer to, he sees exactly what he described in Chapter 13 of Leviathan: a condition in which the most ruthless actors win, and the rest of humanity lives inside the world those actors construct. The EU’s two-year delay on AI Act compliance is, for Hobbes, not a sign that regulation is difficult. It is a sign that the sovereign waited too long, ceded ground it will not easily recover, and is now managing a retreat rather than establishing authority.
There is also an indirect historical antagonism worth noting. Smith was shaped by the Scottish Enlightenment tradition of natural jurisprudence, which was built explicitly in opposition to Hobbes’s account of human nature. Smith’s teacher Francis Hutcheson constructed his entire moral philosophy around the argument that Hobbes was wrong about human motivation. Smith inherited that dispute and carried it into economics, and the Hobbesian shadow is present throughout The Theory of Moral Sentiments, where Smith is persistently arguing against the position that self-interest is the only genuine motive people have.
Who Else We Considered
John Stuart Mill vs Plato was the original pairing we considered for this topic, and it remains strong. Plato’s philosopher-kings as the model for AI governance against Mill’s harm principle as the ceiling for any legitimate regulation would produce a genuinely interesting debate. We held it back because Smith and Hobbes offers something that pairing does not: a direct confrontation between an economic and a political framework. Smith’s argument is not primarily about liberty. It is about incentives and institutional design, which felt more immediately applicable to the frameworks currently being written.
John Locke was considered for the anti-regulation side because his labor theory of property creates genuinely thorny problems for AI: who owns the model, who owns the training data, who owns the output? We passed on him because Smith’s regulatory capture argument maps more directly onto what is actually happening in Washington right now, and Locke would have required more setup to get there.
Alexis de Tocqueville was considered as an alternative to Hobbes. His concept of soft despotism—the form of power that does not tyrannize but infantilizes, that does not forbid but makes alternatives gradually unthinkable—maps almost perfectly onto AI behavioral nudging and algorithmic content shaping. We kept him in reserve because this debate needed Hobbes’s raw confrontational energy. A Tocqueville debate on this topic is still very much on the table.
Woodrow Wilson was briefly considered as the technocratic progressive voice for regulation. Wilson believed governance should be handled by trained experts insulated from democratic pressure, which describes most proposed AI regulatory agency structures precisely. We passed on him because Wilson’s earnestness, while historically authentic, produces less interesting friction than Hobbes. Wilson would sincerely believe in the regulatory agency. Hobbes finds it imperfect and argues for it anyway, which is a more honest and more compelling position.
Why Each Man Takes the Position He Does
Smith’s position flows from Book IV of The Wealth of Nations, where he develops the regulatory capture argument with precision that looks prophetic from 2026. His core claim is that producers in any market dominate consumers in the political process, because producers are concentrated and motivated by large individual stakes while consumers are dispersed and individually motivated by stakes too small to justify sustained political action. The White House framework, written after extensive industry consultation and explicitly recommending industry-led standards over any new regulatory body, is exactly what Smith would predict: the regulated writing the regulation, framed as innovation protection but functioning as incumbent protection.
His support for antitrust rather than comprehensive regulation comes from his analysis of monopoly in Book I and his theory that markets produce good outcomes only under genuinely competitive conditions. Smith would look at the current AI landscape, dominated by a small number of companies with enormous capital advantages and network effects that make competition structurally difficult, and argue that the correct instrument is breaking up that concentration rather than layering regulatory compliance costs on top of it. The White House framework’s recommendation to preempt state AI laws while relying on industry standards would strike Smith as completing the capture rather than checking it.
Hobbes’s position comes from Leviathan Chapter 13, where he describes the state of nature as the condition in which there is no place for industry, no arts, no letters, no society. Applied to ungoverned AI development, this is not metaphorical. Private actors are already deploying AI in ways that erode democratic information environments, automate fraud at previously impossible scale, and concentrate economic power without the social contract obligations that political authority carries. Hobbes’s argument is that the sovereign must establish authority over this technology before power consolidates in hands that the sovereign cannot reach, because once that consolidation occurs, the social contract that makes any other governance possible is already compromised.
The debate in Part Two surfaces three arguments that push both men into uncomfortable territory. The democratic legitimacy problem: if AI can corrupt democratic processes, the sovereign’s authority is undermined by the very thing it is supposed to govern. Hobbes answers that this makes urgency more important, not less. The labor displacement argument: Smith concedes that markets will not adjust fast enough to prevent serious harm, and his disagreement with Hobbes shifts from whether to intervene to which instrument to use. And the China argument, which is where the debate breaks down entirely: international strategic competition in AI does not operate inside the domestic legal structure that Smith’s framework assumes, and Hobbes’s case for sovereign coordination becomes very hard to dismiss when a major state power is directing AI development with the full authority of the sovereign toward strategic objectives.
A Note on the Sources
The primary sources for Smith’s positions are The Wealth of Nations (1776), particularly Books I and IV, and The Theory of Moral Sentiments (1759). His voice is careful, precise, and slightly melancholy: the voice of a man who understands human nature well enough to be consistently disappointed but never surprised. The most useful line for this debate comes from Book I, Chapter 10, where he describes merchants of the same trade meeting together and conspiring against the public. It has been cited so often it has become a cliche, but it remains the most exact description of regulatory capture ever written, and it applies to the March 2026 White House framework with no translation required.
The primary sources for Hobbes’s positions are Leviathan (1651), particularly Parts I and II, and De Cive (1642), an earlier and more compressed version of the same argument. His voice is thunderous and confident: the voice of a man who believes he has identified the one truth that matters and is not interested in being diplomatic about it. One historical note worth acknowledging: Hobbes’s mathematical proofs were publicly demolished by John Wallis in a dispute that lasted decades, which makes his certainty somewhat ironic in retrospect. The logical structure of his political argument holds up considerably better than his geometry did.
Both men are being extrapolated here. Neither had any concept of artificial intelligence, and both wrote in contexts that differ from 2026 in ways that matter. We have tried to honor the intellectual core of each position while being honest that we are applying frameworks, not quoting conclusions.
What Comes Next
The democratic legitimacy question that surfaces in Part Two deserves its own full debate, and we have a pairing in mind that we think will run considerably hotter than this one. More on that soon.
This debate was created using AITalkerApp.com, the tool that turns voice recordings, photos, and scripts into animated conversation videos. If you have a dialogue you want to bring to life, that is where to start.