Who Owns Humanity’s Intelligence?

I recently heard an idea from Russian political writer and commentator Stanislav Belkovsky that stayed with me.

He was talking about the way people fear artificial intelligence as if it were some alien force descending on Earth — reptilian invaders from a distant planet, arriving to take control of humanity.

But artificial intelligence did not come from outer space.

Behind it are very real people: the owners of technology companies, computational infrastructure, and AI models. They gained access to humanity’s accumulated knowledge, turned that knowledge into functioning systems, and now sell humanity access to those systems.

That led me to a different question.

If artificial intelligence is built on knowledge created by humanity over centuries and millennia, who should own the economic value created from its commercial use?

We often talk about AI as if it were already an independent economic and political actor. AI “takes jobs.” AI “makes decisions.” AI “becomes too powerful.”

But a model does not set the subscription price. It does not build the data center. It does not choose the business model. It does not sign the layoff notice.

People do.

So one of the defining conflicts of the AI era may not ultimately be between humans and machines. It may be between different groups of humans: those who own the new productive infrastructure, those who use it to increase profits, and those whose collective inheritance made that infrastructure possible in the first place.

Modern AI exists at the intersection of two very different forms of capital.

The first is private capital: GPUs, data centers, energy, models, engineering teams, research, software infrastructure, and billions of dollars in investment. That capital has owners. It has contracts, equity, balance sheets, and a clear economic price. The people and companies that create it are entitled to earn a return.

But there is another form of capital.

It was not created over ten years, or twenty, or even a century. It was accumulated over thousands of years: language, mathematics, scientific discoveries, literature, history, professional knowledge, software code, art, public debate, and countless ways of describing human experience.

Call it civilizational capital.

It has no single creator. It cannot be placed on the balance sheet of one company. Much of it existed long before modern copyright law, and long before knowledge itself was treated as something that could be formally owned.

This is where today’s copyright debate captures only part of the problem.

Of course there are books, articles, images, programs, and other works whose rights belong to specific people and organizations. The debate over whether and how those works can be used in model training is necessary.

But even if every licensing dispute were resolved tomorrow, an enormous layer of knowledge would remain that cannot be assigned neatly to individual owners: language itself, cultural structures, logic, foundational scientific ideas, public-domain knowledge, and the accumulated experience of generations.

That resource is so old, so embedded in everyday life, that we barely notice its economic value.

Air is not registered in the name of every person who breathes it. Language does not belong to the company that first discovers how to commercialize it efficiently. No one owns a patent on the human capacity to pass experience from one generation to the next.

Some resources exist before the legal systems that later define ownership over more specific things.

Before modern AI, accumulated human knowledge was in some ways like air. It surrounded civilization, enabled its development, and passed from generation to generation, but it did not exist as one scalable commercial asset.

Artificial intelligence changes that.

It creates a machine capable of transforming vast amounts of human knowledge into a new productive force that can be sold to millions of people at once.

At that point, the question moves beyond copyright.

If a private company commercially exploits civilizational capital, should part of the value created from that resource return to its inheritors?

Consider a familiar situation involving a natural resource.

A person owns land. Oil is discovered beneath it. An oil company arrives with engineers, drilling equipment, capital, and technology. It extracts the resource, refines it, and sells products with far more value than the oil had underground.

The company deserves a return on everything it created.

But the landowner receives royalties.

The landowner does not receive gasoline for free. They still buy the finished product at market price. Paying for the final product and receiving a share of the value of the underlying resource are not contradictory.

Why should the economics of artificial intelligence be entirely different?

I personally pay far more for ChatGPT than the average user because I use ChatGPT and Codex intensively. I am paying for compute, infrastructure, research, product development, and access to an extraordinarily capable technology.

That is entirely reasonable.

But if part of the value of that technology comes from humanity’s collective intellectual inheritance, then society has a legitimate claim not to the product itself, and not to ownership of the company, but to a small share of the economic return created from the common resource.

That leads to the idea of a civilizational dividend.

This is fundamentally different from welfare or Universal Basic Income.

UBI is usually framed around need. The state pays people a basic amount because they require a minimum income, especially in scenarios where AI and robotics significantly reduce demand for human labor.

A civilizational dividend rests on a different principle.

A person receives it not because they are poor, unemployed, or economically obsolete. They receive it because they are an inheritor of a common asset.

That distinction matters psychologically as well as economically.

A benefit says: society supports you because you are in a weaker position.

A dividend says: you have a legitimate claim on the returns generated by an asset you partially inherit.

Future generations do not need to prove that they personally wrote books, discovered equations, or contributed data to a training set. They inherit the civilization built by earlier generations. A child is born into a world containing language, institutions, universities, roads, medical knowledge, scientific discoveries, and accumulated cultural capital created long before that child existed.

The intellectual inheritance of civilization should be understood in the same way.

The United States creates an American AI Sovereign Fund.

Large commercial owners of foundation models and other major AI infrastructure contribute a small civilizational royalty on the commercial exploitation of systems built on large bodies of human knowledge.

The money does not disappear into the federal budget. That condition is essential.

The fund exists as a separate long-term asset with independent governance, protected principal, and a transparent distribution formula. Part of the proceeds is paid to citizens. Part is invested and preserved for future generations.

The broad logic is already familiar from sovereign wealth funds and natural-resource funds. What changes is not the financial mechanism but the underlying resource. Instead of oil or gas, the fund recognizes the commercial exploitation of collective intellectual inheritance.

No one can honestly predict the size of the AI economy thirty years from now, so the numbers below are not forecasts. They are scenario ranges designed to make the scale easier to understand.

Suppose that in a mature AI economy, the commercial value directly attributable to AI in the United States reaches $2 trillion per year. A 2 percent royalty produces $40 billion for the fund. With a population of roughly 370 million people, that equals about $108 per person per year.

At $10 trillion in AI-generated commercial value, the same 2 percent mechanism produces $200 billion, or roughly $540 per person.

At $25 trillion, the fund receives $500 billion, or approximately $1,350 per citizen.

At a 5 percent royalty, those same scenarios produce roughly $270, $1,350, and $3,380 per person per year.

These numbers are not predictions. We do not know the future size of the AI economy. We do not know the optimal royalty rate. We do not even know exactly which forms of revenue should qualify as commercial exploitation of civilizational capital.

The point is simply to establish an order of magnitude.

And the calculation reveals something important.

A civilizational dividend can become meaningful additional income and, over time, build a very large public asset. But under plausible assumptions, it does not replace a salary.

If a person receives several hundred or several thousand dollars per year, that improves financial resilience. But if that person loses a job paying $70,000 or $100,000 a year, the dividend is not enough.

So AI-driven labor displacement requires a separate mechanism.

People lose jobs for many reasons. A business fails. Demand collapses. Management changes. An employee commits serious misconduct or repeatedly violates workplace rules.

There is no reason to treat all of these cases as AI displacement.

But consider a different situation.

A company deploys an AI system, eliminates a group of positions, and then produces roughly the same amount of work with fewer people.

The company saves money on payroll and benefits. The worker loses income. Government and society absorb unemployment costs, retraining expenses, lower tax revenue, and the broader social effects of displacement.

Part of the real cost of automation has moved outside the corporate balance sheet.

That is one reason automation can appear economically cheaper than it actually is for society as a whole.

This is where a second mechanism enters: the Labor Replacement Levy.

It is not a tax on AI itself.

A company uses AI to help employees work better, increase output, improve services, or create new jobs without reducing its workforce. No special obligation arises.

The obligation begins when there is verified labor replacement.

If part of the economic benefit comes directly from eliminating jobs, a portion of that benefit helps finance the transition of the people who lost their income.

That support can take several forms: a supplement to existing unemployment insurance, a dedicated training account, temporary wage-loss insurance if the worker finds a lower-paying job, or bridge support for workers close to retirement.

The system also has to avoid punishing employers twice.

If a company already provides meaningful severance, pays for retraining, extends health coverage, or successfully moves workers into new roles, those costs count toward its obligation.

The goal is not to collect as much money as possible.

The goal is to prevent a company from transferring the transition cost of its own automation decision entirely onto the worker and society.

The Labor Replacement Levy has another function that may be even more important than compensation.

It becomes a regulator of the speed of AI deployment.

The debate over AI development often falls into two camps. One side wants technology to advance as quickly as possible. The other looks for ways to slow it down through regulation, international agreements, voluntary pauses, or limits on the most powerful systems.

But stopping a global technological race through agreement is extraordinarily difficult. If one country deliberately slows fundamental research while a competitor continues accelerating, the result can be strategic disadvantage rather than global safety.

That suggests a different target for regulation.

Instead of trying to regulate the creation of intelligence itself, we can regulate the speed at which technological capability becomes mass social displacement.

A company can deploy new software in months. Society cannot create millions of new careers, restructure universities, retrain adults, and restore their previous earning power on the same timeline.

Every economy therefore has a limit to how quickly it can adapt.

Call this the Social Absorption Rate.

It is the speed at which a labor market can absorb people displaced by technology: create new roles, retrain workers, and return them to economic independence.

When AI displaces labor more slowly than this rate, society adjusts. Some jobs disappear. Others emerge. Productivity rises.

When displacement moves faster, the gap becomes long-term unemployment, declining wages, weakened local economies, and political instability.

The Labor Replacement Levy puts a price on that gap.

A company remains free to automate. But the business calculation is no longer simply the cost of AI minus the salaries of the people being removed. It now includes part of the cost of helping those people transition.

If automation creates enormous productivity gains, it still makes economic sense.

If the business case works only because large social costs are pushed onto everyone else, it becomes less attractive.

The result is an economic brake that does not ban AI or stop research. It synchronizes the speed of labor replacement with society’s capacity to adapt.

This is the obvious objection.

If the United States makes automation more expensive and China does not, do Chinese companies gain an advantage?

Yes — if the American system simply imposes a large tax on every use of AI. That would create a genuine competitiveness problem.

But the Labor Replacement Levy works differently.

It does not penalize model development. It does not make every AI feature more expensive. It appears only when technology directly translates into the elimination of jobs.

And China is not constrained only by the number of GPUs it owns or the speed at which it can deploy robots. It is constrained by society too.

Even a country with more than a billion people cannot replace human labor indefinitely without confronting the question of where people earn income, what economic role they occupy, and how domestic demand is sustained.

A highly automated factory can produce extraordinary volumes of goods. But an economy still depends on people who can buy those goods, build families, pay taxes, and believe that the economic system offers them a future.

The international AI race therefore does not eliminate the Social Absorption Rate. It merely makes the problem politically harder.

Outsourcing, offshoring, international competition, and possible ways companies might avoid a Labor Replacement Levy are large questions in their own right. They deserve a separate analysis.

But difficult edge cases do not erase the underlying reality: technological speed and social speed are not the same thing.

At this point, the system is still incomplete.

The civilizational dividend gives people a claim on the common inheritance.

The Labor Replacement Levy protects them when technological displacement destroys an existing source of income.

But both mechanisms are still, ultimately, payments.

They protect people from loss. They do not necessarily make people participants in the upside of the new economy.

That requires a third mechanism: Distributed AI Capital.

The idea of distributed AI infrastructure already exists in various forms. David and Daniil Liberman, for example, have argued for broader and more distributed ownership of the computational infrastructure underlying AI. But the principle can extend far beyond GPUs and compute networks.

The AI economy is creating an entirely new class of productive assets: computational capacity, autonomous agents, specialized AI systems, robotic enterprises, energy infrastructure for large-scale compute, and companies in which a growing share of productivity comes from machines.

If most of those assets remain concentrated in the hands of a small number of corporations and wealthy capital owners, automation almost inevitably increases wealth concentration. Income shifts from labor toward capital while most people remain dependent on wages — or eventually on transfers.

But that is not the only possible structure.

Distributed AI Capital means broadening ownership of the productive assets of the AI economy.

This can take many forms.

Pension funds own stakes in AI infrastructure. Citizen investment funds finance robotic enterprises. Employees receive ownership interests in the automated productive assets of their own companies. Local communities invest in energy and compute infrastructure. Individual investors gain access not only to shares of the largest technology corporations but also to more distributed productive AI assets.

The central idea is not the precise legal vehicle.

It is the transition from redistributing income after value has already been created to distributing ownership before the income is created.

A person who exists economically only as a worker is naturally placed in tension with automation. The more productive the machine becomes, the more vulnerable the worker’s bargaining power may become.

A person who both works and owns part of the new productive infrastructure occupies a different position.

The machine may change that person’s profession. But rising productivity also increases the value of assets they own.

They are no longer only competing with the technology.

They are becoming its co-owner.

Only now does the full structure come together.

Distributed AI Capital gives people private income from ownership of the new productive infrastructure.

The civilizational dividend provides public ownership income derived from humanity’s shared intellectual inheritance.

The Labor Replacement Levy protects people during the period when technological displacement moves faster than they can find a new economic role.

These are three different mechanisms because they solve three different problems.

Distributed capital answers: who owns the productive force of the future?

The civilizational dividend answers: who has a legitimate claim on value created from humanity’s shared intellectual inheritance?

The Labor Replacement Levy answers: who pays for the transition when corporate automation destroys an existing source of human income?

Together they form an economic architecture for the AI transition.

A person continues earning wages while their labor remains valuable. At the same time, they earn investment income if they own part of productive AI capital. As a citizen and inheritor of civilizational capital, they receive a permanent dividend. If technological change destroys their profession, temporary transition insurance gives them time to retrain, move into another field, or bridge the remaining years to retirement without financial collapse.

None of these income sources has to be sufficient on its own.

Their combination is what changes a person’s economic resilience.

The most important effect of this system may never appear in a spreadsheet.

It is confidence in the future.

Economies run not only on productivity, technology, and capital. They also run on expectations.

People buy homes because they believe they will still be able to pay the mortgage ten years from now. They have children because they believe they can support a family in the future. They start businesses because risk does not mean immediate financial ruin. They change careers because they believe they can survive the transition.

A person who sees AI only as a threat to their job has every reason to resist automation.

A person who owns part of the new economy, receives a dividend from shared civilizational inheritance, and knows that sudden displacement will not leave them without support experiences technological change differently.

That creates another productive resource: economic confidence in the future.

It can reduce resistance to change. It can make retraining easier to accept. It can encourage entrepreneurship. It can preserve consumer demand during structural transformation. And it may reduce the political radicalization that emerges when large parts of a population become convinced that the future is being built at their expense.

That is why the economics of AI cannot be reduced to productivity alone.

Productivity determines how much wealth society creates.

Ownership determines who receives that wealth.

The transition system determines how many people are damaged along the way.

One distinction is essential across all three mechanisms.

Technology companies are not the enemy.

They have built systems that did not exist before. They have invested enormous amounts of capital, taken extraordinary risks, assembled some of the world’s strongest engineering teams, and created infrastructure capable of turning knowledge into a new productive force.

They are entitled to profits.

Entrepreneurs are entitled to profits.

Investors are entitled to returns.

People who create better technologies should receive an economic reward for success.

But private capital is not the only source of the value being created.

AI emerges inside human civilization and draws on its accumulated intellectual inheritance. Automation takes place inside societies and creates costs that do not disappear simply because they do not appear on one company’s income statement. And the new productive infrastructure does not have to belong exclusively to a small group of owners.

The proposed system therefore does not abolish markets.

It tries to make them more complete.

Private capital earns returns for building the technology.

Citizens receive a civilizational dividend for the commercial use of the common inheritance.

People can own Distributed AI Capital and participate in the upside of the new economy.

Employers finance part of the transition when their decisions directly destroy human income.

This is not an attempt to equalize outcomes.

It is an attempt to recognize different rights, different contributions, and different forms of responsibility.

We are accustomed to asking when artificial intelligence becomes smarter than humans, how many professions it will eliminate, whether AGI will arrive, and whether one day we will lose control over a system we created ourselves.

Those questions matter.

But between today’s world and any possible world of superintelligence lies a far more immediate problem.

Who owns the productive power of AI?

Who receives the income generated from humanity’s collective intellectual inheritance?

Who pays for the transition when professions disappear?

And how quickly can society reorganize itself without destabilizing the lives of millions of people?

Artificial intelligence did not arrive from another planet.

It grew inside human civilization — from its language, memory, knowledge, mistakes, and discoveries.

So the central question is not whether AI takes over our world.

The central question is what institutions we build around it.

We can create an economy in which a small group of owners controls the new productive infrastructure, most people remain dependent on wages, and those who lose them eventually become dependent on transfers.

Or we can build a different architecture.

One in which people remain workers while their labor is valuable, become owners of part of the new productive force, receive dividends from the civilization they inherited, and gain time and resources to adapt when technology destroys old work faster than the economy can create new roles.

That is not a world in which governments are forced to support people after machines have “won.”

It is a world in which people retain an economic stake in the civilization they built.