In the spring of 1932, the richest man in American electricity boarded a boat for Europe with creditors at his back and the first indictments forming behind him. Samuel Insull had spent three decades building a utility empire that reached millions of customers across the Midwest and beyond, a lattice of operating companies and holding companies stacked one atop another until the structure was tall enough to light a good part of the country and fragile enough to fall in a single season. He had also spent those decades as the most fluent evangelist of private power in America. Government had no business generating electricity. Public ownership would frighten off capital, smother enterprise, and hand the lights to bureaucrats who could not run a streetcar. The discipline of the market was the one thing standing between the consumer and ruin.

The empire collapsed that year, and public competition had nothing to do with it. The arithmetic underneath the holding companies had always been a confidence trick, and the Depression was the gust of wind that set the house of cards folding. Insull fled, was chased across the Mediterranean, extradited, and tried for fraud. A jury acquitted him. He died a few years later in a Paris subway station with a few francs in his pocket and an estate that owed millions (McDonald, 1962).

Hold Insull and his story in your mind for a moment. We will come back to him, because the argument he made in his prime is being made again, almost word for word, by men who have thousandfold his fortunes, but have not yet gotten to the fleeing part of the story quite yet.

About a week ago, Bernie Sanders used the opinion page of The New York Times to announce a bill he had not yet written. The American A.I. Sovereign Wealth Fund Act, he said, would arrive in the coming weeks, and it would do something no American law has done to a major industry in living memory. It would take half of it (Sanders, 2026). Not half the profits. Half the stock. A one-time levy of fifty percent of the equity in OpenAI, Anthropic, xAI, and the other large model-makers, paid in shares rather than cash, deposited into a public fund that would carry voting rights, hold an equal seat on each company’s board, pay dividends to ordinary Americans in the near term, and underwrite health care, education, and housing as it grew (Sanders, 2026).

Sanders’ core justification rests on a claim about where the value of current LLM’s came from. The models were trained on the collected output of humanity: our books, our music, our code, our journalism, our recorded speech, our images, the accumulated labor and invention of generations, much of it ingested without permission, acknowledgment, or payment (Sanders, 2026). Sam Altman has said as much himself, describing the training data as the collective experience and knowledge of mankind (Sanders, 2026). Sanders draws the obvious inference. A fortune built on a public inheritance owes the public a share.

The men who built the models have been less delicate about that inheritance than Altman’s phrasing lets on. In a 2024 interview, Mark Zuckerberg said as much without flinching, telling his interviewer that creators and publishers “overestimate the value of their specific content,” and that anyone who refused to be scraped would simply be left out, since their absence would not change the outcome much (Robertson, 2024). The disposition is plain: the work was worth taking, and the people who made it were worth nothing. The taking, by one account now before a court, ran past the open web into deliberate theft. In May 2026 five publishers, among them Hachette, Macmillan, and McGraw Hill, joined the novelist Scott Turow in a class action in the Southern District of New York alleging that Meta torrented millions of copyrighted books and journal articles from the pirate libraries LibGen and Anna’s Archive to train its Llama models, that Zuckerberg personally authorized it, and that the company chose piracy over licensing on purpose, since paying for even one book would have weakened the fair-use defense it intended to claim (NPR, 2026). Meta denies wrongdoing and rests on fair use, and a separate authors’ suit on similar ground failed in 2025 (Variety, 2026). The piracy is an allegation a court will weigh. The contempt is not in dispute. It was entered into the record in 2024, in his own words.

The remarkable thing is how much of this ground the industry has already ceded. OpenAI has floated a public wealth fund that would give every citizen a stake in AI-driven growth. Anthropic has endorsed national sovereign wealth funds holding positions in AI. Elon Musk, who runs one of the targeted companies, has called for the federal government to mail checks to citizens displaced by the technology (Sanders, 2026). The disagreement that remains is narrower than it looks. It is a fight over how large the share should be and who holds it.

Which is why the reaction to Sanders has been so revealing, and so old. One outlet ran its verdict in the headline, declaring that the senator understood neither AI nor wealth (Reason, 2026). A column at PJ Media asked whether he was nuts, called the plan confiscation, and wondered aloud whether any government could be trusted to run such a fund for the public good rather than its own (Moran, 2026). Market analysts pronounced the mechanism legally and politically impossible (Newsweek, 2026). Strip the AI vocabulary out of those objections and what remains is Insull’s stump speech from 1931. Confiscation will scare off the capital. Government cannot run the thing. It is socialism, and it will kill the goose.

The forecast has a long bibliography and a losing record.

It also arrives in 2026 already contradicted by the administration its loudest messengers support. In February 2025 the Trump White House issued an executive order to establish an American sovereign wealth fund of its own (The White House, 2025). The federal government already holds equity in private firms, having taken positions in Intel and the rare-earth miner MP Materials in the months before Sanders spoke (Fortune, 2026). The charge that public ownership of a strategic industry is foreign to the American system is being leveled by people who voted for the man who just did it. The objection tracks whose interest is at stake, and it always has.

Sanders’ deeper claim, the one about the public having already paid, carries a pedigree he never bothers to spell out, but which both the internet as we know it and AI are dependent upon. The internet was not conjured in a garage. It began as ARPANET, built with Defense Department money in 1969, scaled through the National Science Foundation’s NSFNET in the 1980s, furnished with its foundational research by public universities and its training corpus by the open web (Abbate, 1999). The economist Mariana Mazzucato has spent a career documenting the pattern. The state funds the dangerous, patient, foundational work, and private firms arrive later to harvest the returns and write the public hand out of the founding myth (Mazzucato, 2013). The touchscreen, the satellite positioning, the voice recognition, the protocols themselves were seeded with public dollars and capitalized into private fortunes.

Here is the part of that genealogical tree that should make a careful advocate uneasy. The public’s contribution never came with shares. Washington paid to invent the network and owned no part of the companies that grew on top of it. The moral claim is airtight. The equity claim, judged as historical precedent, is thin. Sanders is asking the country to convert a paternity it never monetized into an ownership it never held previously.

There are precedents for that conversion, and they are worth examining for what their opponents predicted and what came to pass instead. Different times, different technologies, but a very similar pattern.

Begin with the closest American analogue to a public stake in a transformative network. In 1933 the Tennessee Valley Authority started generating public power in a region the private utilities had surveyed and abandoned as unprofitable. Two years later the Rural Electrification Administration began wiring the farms the power trust would not touch. The governing idea was the yardstick. Public power would stand as a measuring stick against which private rates could be judged and disciplined, a benchmark the monopolies could not slip out from under. The utilities fought it as unconstitutional socialism and carried that fight to the Supreme Court, which upheld the Authority’s power to generate and sell electricity (Ashwander v. Tennessee Valley Authority, 1936).

What the prophecy got wrong is the crux of the story. Rural electrification took a countryside where roughly one farm in ten had power and, within a generation, left nearly all of them lit. Those new wires created a continental appetite for motors, refrigerators, radios, and milking machines, an appetite private manufacturers raced one another to feed. The socialism that was supposed to strangle American industry built one of the largest consumer markets in human history. And the man who had been the loudest voice on the other side, who swore public power would wreck the country, had already wrecked himself by then, in a manner that owed nothing to public competition and everything to his own books.

That detail is heavier than it looks. This metaphorical yardstick is exactly what Sanders reaches for when he says the public’s board seats would let it block decisions that harm citizens (Sanders, 2026). A public stake works as a standing discipline on the private actor, a measurement the monopoly cannot escape. The dividend is the visible benefit. The discipline is the one that frightens the incumbents far more than the check in the mail.

The question underneath all of this: Who should own the wealth a new machine age throws off,? This issue is older than any of the technological pathways under discussion. As I wrote about last week , in 1891, at the high tide of the first industrial revolution, Leo XIII wrote Rerum novarum, the encyclical that founded modern Catholic social teaching, and he refused both of the easy answers. He condemned a capitalism that had laid on the laboring poor a yoke little better than slavery, and in the same breath he rejected the socialist cure of abolishing private property, which he judged would rob the worker of the fruit of his own labor and deliver too much into the hands of the state (Leo XIII, 1891). The just arrangement lived in the narrow country between the rapacious market and the confiscatory state, and finding that country was the whole intent of the document. When Cardinal Prevost took the name Leo XIV in 2025, he said he chose it because Leo XIII had answered one industrial revolution and the Church now faced another in artificial intelligence, the same questions about human dignity and labor walking back in under a new name (Yang, 2025). The needle Leo XIII threaded is the needle every public-ownership fight has to thread, and it is the needle the Sanders proposal will live or die on.

Norway is the case Sanders cites, in his writing, and the case his opponents ought to study hardest (Sanders, 2026). When the North Sea oil came in, the foreign majors warned that state ownership would drive off the deepwater expertise Norway did not have and leave the wealth in the seabed. Norway licensed the majors, wrote rules to keep the rents at home, founded a national oil company, and built its own competence anyway. The sovereign fund that grew out of that decision now holds more than two trillion dollars and owns a sliver of nearly every public company on earth (Sanders, 2026; Fortune, 2026). The prophecy that public ownership would shrink the pie was simply false. The pie grew, and Norway cut slices for everyone.

It is not all rosy tales, there are some real examples of awful outcomes.  In another country, another hemisphere. Venezuela nationalized its oil and, over years of running the national company as a patronage machine and a piggy bank, hollowed it until production cratered. Britain took the same instrument and produced British Leyland, a nationalized carmaker that became a synonym for stagnation. Reach back further and the American transcontinental railroad, built on public land grants and public bonds, was looted from the inside through the Crédit Mobilier scheme, public money turned to private plunder under cover of a public work. These are the ballast that keeps the case honest. Public ownership does not reliably produce Norway. The asset is amoral, but  the institution (and politicians) holding it decides whether it becomes a pension or a heist.

What Sanders is stating, and what I am assuming, is that AI is the kind of thing a fund can hold the way Norway holds oil.  I tend to think that the Trump administration is looking at things from a decidedly more Venezuelan perspective, which, given its recent activities in Venezuela, are too thick with irony to properly address as a sidebar.  At any rate, there are two serious objections to the assumption that the US could turn AI into a trust like Norway did with the north sea, and those objections point in opposite directions.

The first objection is hype. In AI Snake Oil, Arvind Narayanan and Sayash Kapoor argue that the technology’s capabilities are routinely and systematically overstated, and they borrow Lee Vinsel’s term criti-hype for the way even the loudest alarms about AI end up inflating it, selling the product by dressing it in apocalypse (Narayanan & Kapoor, 2024). Sanders opens his own case by calling AI almost certainly the most transformational technology in the history of the world (Sanders, 2026), which is the labs’ marketing copy recited back by their antagonist. You can build a fund around a seabed of oil because the oil is there and can be measured. A valuation is a softer thing to nationalize, and a bubble is a hard thing to put in a vault.

The second objection is control, and it cuts deeper. Roman Yampolskiy argues that it is unwarranted to say anyone owns an advanced AI system at all, on the grounds that no one controls it: not its code, not its outputs, not its goals, not the data it was built from (Yampolskiy, 2024). Norway could send a rig to a coordinate and order it to drill. A federal board cannot send an instruction into a system its own creators cannot fully explain. Owning the equity and steering the technology are two different acts, and a share certificate makes a poor rudder, particularly as techbros start to toy with launching LLM-based business ventures that are run entirely by LLM’s.

Take both objections at full strength, and watch what they do.

Grant the hype case completely. A valuation inflated past the real worth of the thing is a speculative claim staked on knowledge the public already owns. Socializing that claim puts the risk of the bubble onto the bubble and leaves the public holding a brake on the mania rather than a sack of it. The cheaper the asset turns out to be, the cheaper the public got in, and the more the whole exercise reads as discipline imposed on a frenzy. The hype argument and the public argument arrive at the same door.

Grant the control case completely. A technology no one can steer is the most dangerous possible thing to leave in the hands of a few men who answer to no one. The observation that a public board seat might be inert against a self-improving system is also an observation that a private board seat is inert, which means the men holding those seats today are steering nothing while telling the world they have it well in hand. Uncontrollability is an argument for more hands on the wheel and more eyes in the room. The case for public governance strengthens precisely as the case for anyone’s governance weakens.

What the skeptics actually win is a mechanism to blunt the impact that these companies are already having, and it is the most valuable thing they could win. A fund that holds half of the most valuable companies in the country has to be built the way Norway built its. Insulated from the appropriations cycle. Run by professionals. Walled off from every politician who will eventually want to raid it to paper over a budget hole. Built carelessly, it does not become Norway. It becomes Venezuela with data centers, or worse, Crédit Mobilier with a board seat. The history is unanimous on this single point. The idea stays constant across every one of these cases. The institution is what varies, and the governance of the institution is what decides.

Set the mechanism aside for a moment and look at the thing it is for, because the dividend is the smallest part of it.

The oldest name we have for wealth that sits in one place and feeds no one is the dragon. Fáfnir was a dwarf before he was a worm. He took the cursed hoard of Andvari, coiled himself around it, and the gold turned him into the monster that guarded it. The myth is exact about the order of operations. The hoard came first. The dragon was what the hoard made of the one who would not let it go. That is the part the word psychopath gets wrong when it is thrown at the billionaires at the helm of AI initiatives. It puts the pathology in the man, where it is hard to prove and easy to deny. The pathology actually lives in the arrangements they have set up. Joel Bakan made the case plainly: the modern corporation is built to pursue its own advantage with no regard for anyone outside itself, and an entity built that way, set against the clinical checklist, behaves like a psychopath, with the form mandating the conduct regardless of who staffs it (Bakan, 2004).

These AI dragons sit on hoards of their own. What they are guarding is a commons. The training corpus is the accumulated expression of nearly everyone who ever wrote, drew, sang, argued, or recorded a thought where a machine could later absorb it, which means it belongs, in any honest accounting, to the dead and the living together and to no single company. Every teacher who explained a concept clearly enough to be transcribed, every coder who posted a fix, every musician whose phrasing entered the set put a brick in the foundation, and not one of them was asked or paid. John Locke set the condition for taking anything out of the common store: a person may appropriate from it only so long as enough, and as good, is left for everyone else (Locke, 1689). The labs took the whole of it and left nothing as good behind, because there is one such corpus and it has now been enclosed. The writer whose sentences trained the model cannot use those same sentences to train a rival and arrive first. The LLM training methodologies did not create knowledge. They refined a copy of it and put a meter on the original. The fence is up, and the rent is set by the people who built the fence. Strip the futurism off and the ethical shape is ancient: a commons turned into a toll road, with the toll charged back to the people who made the road by walking it.

The reflexive reply is that a commons cannot be held in common without ruin, that shared ownership is a tragedy waiting on its third act. Elinor Ostrom spent a career and earned a Nobel showing this to be false, documenting communities that governed shared resources well and for centuries when the institution holding the resource was built with care (Ostrom, 1990). Sanders’ proposed fund is a wager on that finding. A commons enclosed by a few can be returned to the many by an institution that holds it on their behalf, so long as it is built like the ones that lasted and not like the ones that were looted. The history kept arriving at this point from the side of engineering. Here it arrives from the side of ethics. The asset wants the right keeper.

What citizens get from the right keeper runs deeper than a check. Philip Pettit revived an old republican idea of what freedom is: to be free is to live beyond the reach of another’s arbitrary power, clear of any master who could harm you and merely chooses, today, not to (Pettit, 1997). A small number of men now hold the steering wheel of a technology that will reset what work pays, what information can be trusted, how children are taught, and what a human voice is worth in a market. They may steer it well. Whether they steer it well is, at present, entirely theirs to decide, and a free people does not live at the mercy of anyone’s good intentions. Dependence on a benevolent master is still dependence, and the master’s benevolence is a fact about him rather than a right held by anyone else. The board seat is the line between a public governed by the technology and a public with a hand in governing it.

The arithmetic of risk makes the same case in colder language. The billionaires at the helm of these disruptive tech companies carry a bounded downside. They keep half, they stay rich, and the worst plausible outcome still leaves them incalculably wealthy. The public carries an unbounded downside. The public at large absorbs the displaced worker, the hollowed-out newsroom, the surveilled street, the manipulated election, and the child taught by a machine tuned for engagement rather than understanding. The party bearing the risk holds none of the reward and no part of the wheel. The party holding the reward and the wheel bears almost none of the risk. Musk’s call for the government to mail checks to the unemployed is an admission against interest (Sanders, 2026). It grants that mass displacement is coming and that someone will pay for the cushion. The public stake is only the question of who. A fund built on the equity pays for that cushion out of the same rising value that is creating the need for it, rather than out of a tax on the wreckage after the fact.

There is a difference between a gift and a debt, and it is the whole distance between dependence and standing. The checks Musk and his ilk propose to mail are offered as largesse, the patron’s hand extended down to the displaced, and a hand that gives can always be withdrawn. A share held by right is what is owed, and the one who holds it stands as a creditor rather than a supplicant. Leo XIII drew the same line when he insisted that justice for the worker could not be discharged as charity from the employer, that what is owed in justice is not satisfied by what is offered in kindness (Leo XIII, 1891). The public fund changes the public’s posture from open palm to part owner, and that change of posture is worth more across time than any dividend it pays.

All of this turns on the hoard. The leaders of these companies could be saints and the argument would not move, because the argument was never about them. The one who sits on the hoard always believes two things at once: that it is his by right, and that the village cannot be trusted to hold it. The dragon is always sincere. He is also, across the long run of these stories, almost always wrong about the village, and almost always undone by the gold, or someone the villagers convince to rid them of the dragon.

This brings us back to Insull, the dragon this story started with, dead on a subway platform with his prophecy already disproven by the lights coming on across a countryside his own industry had written off as worthless.

The men issuing the 2026 version of his foreboding warnings lack any of the noble characteristics that separate them from the dragons they would become. What they share with Insull is the forecast, and the vantage it is spoken from. The warning that public ownership will frighten off the talent, that government cannot be trusted with the asset, that the whole thing is confiscation in the costume of fairness, has been spoken in every one of these fights by the parties with the most to lose, and it has been wrong about the outcome nearly every time it has been tested properly. Those warnings were wrong about Norwegian oil. Those warnings were wrong about rural electricity. Those warnings were right, in a way, about Venezuela and Leyland, which Sanders cannot afford to wave off, and which the bill must address in structure.

Leo XIII found his answer in the narrow country between the rapacious market and the confiscatory state, and the whole difficulty was the wayfinding and border drawing. The public will end up holding some share of the machines built on its own inheritance. Trump’s order and the stakes already sitting in Intel and MP Materials have quietly settled that much (The White House, 2025; Fortune, 2026). The open question is the harder one, whether the country can build an institution honest and durable enough to hold that share without turning a yardstick into a heist? It seems doubtful in the hands of the current executive administration, yet, despite that, the legislative branch may be armed here as a means of creating an enormous check to regain balance.

A yardstick in steady hands measures a monopoly and can be turned on that monopoly, in a pinch, to discipline it. The same yardstick, handed to the wrong people, becomes one more thing to steal. Insull spent his life insisting the public could not be trusted to hold the yardstick. He was holding a crooked one the entire time.

References

Abbate, J. (1999). Inventing the internet. MIT Press.

Ashwander v. Tennessee Valley Authority, 297 U.S. 288 (1936).

Bakan, J. (2004). The corporation: The pathological pursuit of profit and power. Free Press.

Fortune. (2026, June 3). Bernie Sanders wants Americans to own a piece of AI. The Trump White House seems to agree. https://fortune.com/2026/06/03/bernie-sanders-ai-ownership-sovereign-wealth-fund-electrification/

Leo XIII. (1891). Rerum novarum [Encyclical letter]. The Holy See. https://www.vatican.va/content/leo-xiii/en/encyclicals/documents/hf_l-xiii_enc_15051891_rerum-novarum.html

Locke, J. (1988). Two treatises of government (P. Laslett, Ed.). Cambridge University Press. (Original work published 1689)

Mazzucato, M. (2013). The entrepreneurial state: Debunking public vs. private sector myths. Anthem Press.

McDonald, F. (1962). Insull. University of Chicago Press.

Moran, R. (2026, June 4). Sen. Sanders wants to take 50% of AI profits to create a sovereign wealth fund. Is he nuts? PJ Media. https://pjmedia.com/rick-moran/2026/06/04/sen-sanders-wants-to-take-50-of-ai-profits-to-create-a-sovereign-wealth-fund-is-he-nuts-n4953572

Narayanan, A., & Kapoor, S. (2024). AI snake oil: What artificial intelligence can do, what it can’t, and how to tell the difference. Princeton University Press.

Newsweek. (2026, June 2). Millions would get “direct payments” under Bernie Sanders’ proposed AI sovereign wealth fund. https://www.newsweek.com/direct-payments-ai-sovereign-wealth-fund-bernie-sanders-12021907

NPR. (2026, May 5). Scott Turow, Macmillan, McGraw Hill sue Meta for AI copyright infringement. https://www.npr.org/2026/05/05/nx-s1-5812623/scott-turow-meta-lawsuit

Ostrom, E. (1990). Governing the commons: The evolution of institutions for collective action. Cambridge University Press.

Pettit, P. (1997). Republicanism: A theory of freedom and government. Oxford University Press.

Reason. (2026, June 2). Bernie Sanders’ AI wealth fund bill shows that he doesn’t understand AI or wealth. https://reason.com/2026/06/02/bernie-sanders-ai-wealth-fund-bill-shows-that-he-doesnt-understand-ai-or-wealth/

Robertson, A. (2024, September 25). Mark Zuckerberg: Creators and publishers ‘overestimate the value’ of their work for training AI. The Verge. https://www.theverge.com/2024/9/25/24254042/mark-zuckerberg-creators-value-ai-meta

Sanders, B. (2026, June 1). The public should own half of the big A.I. companies. The New York Times. https://www.nytimes.com/2026/06/01/opinion/artificial-intelligence-bernie-sanders.html

Variety. (2026, May 5). Mark Zuckerberg “personally authorized and actively encouraged” Meta’s massive copyright infringement to train AI systems, publishers and Scott Turow allege in lawsuit. https://variety.com/2026/digital/news/meta-ai-mark-zuckerberg-copyright-infringement-lawsuit-publishers-scott-turow-1236738383/

The White House. (2025, February 3). A plan for establishing a United States sovereign wealth fund [Presidential action]. https://www.whitehouse.gov/presidential-actions/2025/02/a-plan-for-establishing-a-united-states-sovereign-wealth-fund/

Yampolskiy, R. V. (2024). AI: Unexplainable, unpredictable, uncontrollable. CRC Press.

Yang, A. (2025, May 11). Pope Leo XIV says advancement of AI played a factor in his papal name selection. NBC News. https://www.nbcnews.com/world/the-vatican/pope-leo-ai-played-factor-papal-name-rcna206126