The proposal to seize half of the equity held by leading AI companies would breach the Takings Clause of the Fifth Amendment and would empower the government to oversee a critical industry in ways that mirror certain contemporary policy directions.
A recent New York Times piece spotlighted a plan from Senator Bernie Sanders to have the federal government confiscate half of the stock held by major AI developers. If Congress were to approve such a move, it would run afoul of the Takings Clause of the Constitution.
Sanders defends this expropriation by arguing that AI rests on the “collective knowledge of humanity.”
Artificial intelligence did not emerge from nowhere. The data and language fueling generative A.I. tools were not conjured into existence by Sam Altman or Elon Musk alone. AI stands on the input of our shared intelligence: books, music, art, journalism, software, scientific research, videos, conversations, images, and ideas gathered over generations. That view isn’t merely Bernie Sanders’s; it reflects a broad understanding.
For much of this material, the outputs of tech magnates have been shaped without permission, without acknowledgment, and without fair compensation. Put differently, the creative labor of millions of people—from writers and artists to teachers, scientists, journalists, and ordinary citizens—has effectively been appropriated by some of the world’s wealthiest individuals. The moment has come to reclaim it.
Because AI is rooted in humanity’s collective knowledge, the wealth it generates should ultimately serve the public good.
The Takings Clause of the Fifth Amendment forbids the government from taking private property without providing just compensation. As scholars Richard Epstein and Eduardo Penalver—who disagree on many policy questions—note in a joint essay for the National Constitution Center, the constitutional guarantee applies at minimum to situations involving the outright confiscation of property. Stocks are private property, and taking half the value of stock in major businesses clearly constitutes confiscation.
And the fact that Sanders describes the move as taking only 50% of the stock does not alter the character of the act. If the government seized half of a person’s home or half of a business’s value, that would still be a taking. The Supreme Court has even recognized that a smaller encroachment on property can amount to a taking, as in the landmark case Loretto v. Teleprompter, where the city of New York required a building owner to permit the placement of a cable box on a rooftop. The same principle applies here.
Sanders characterizes the measure as a “one-time 50 percent tax,” but this label has no bearing on its constitutional nature. It remains a direct expropriation of property, not merely a tax on income or a property tax. A fundamental right of property is the power to decide how a resource is used; Sanders’s plan makes the government’s control the central objective. While there can be cases where the boundary between a tax and a taking blurs, this proposal clearly lies on the taking side of that line.
If simply calling an expropriation a tax could shield the government from takings liability, the same tactic could be used to seize almost any asset. The government might take your home by declaring it an in-kind tax paid as land-use rights. It could seize a business or a charity by describing its control over all activities as a one-off tax payable in exchange for governance rights. And so on.
Sanders might attempt to sidestep Takings Clause limits by substituting coercive regulatory pressure, favorable subsidies, or confiscatory export levies on those who resist, rather than outright confiscation. Donald Trump has, in practice, employed similar strategies to gain stakes in firms such as Intel, and there are discussions about applying comparable tactics to major AI companies. The legal objections to the Trump-style approach are real, but they are arguably less straightforward than those to Sanders’s explicit expropriation plan.
Beyond constitutional issues, the Sanders plan—like Trump’s analogous proposals, which I have criticized—is morally and practically flawed. The argument rests on AI’s supposed basis in “collective knowledge of humanity,” but such reasoning could justify taking almost anything. Virtually every productive endeavor relies to some degree on knowledge built by others. Your home, your phone, your automobile, and even your refrigerator all reflect prior discoveries and designs. Anyone who writes a book or article builds on centuries of ideas. Even my own arguments about democratic theory draw on ancient roots of democracy.
AI developers, like many others, stand on a foundation of accumulated knowledge, yet they also contribute significant innovations. The government has no rightful claim to expropriate them. Market competition and consumer choice—not state fiat—should determine the value ascribed to AI products, not political decree.
If concerns exist about potential IP violations—where creators’ works may have been used without authorization—the proper remedy lies in litigation seeking damages, not confiscation. Several such lawsuits are already in progress. A federal expropriation of AI firms would not compensate those whose intellectual property was misused; it would simply redirect any illicit gains to the government.
Sanders also argues for government control over AI because it is a strategically important technology that should not fall under the sway of a handful of billionaires. Yet a long history of socialist experiments demonstrates that state ownership of major industries tends to yield misery—poverty, repression, and tragedy. And, as I have argued in detail elsewhere, Sanders’s version of “democratic socialism” is unlikely to do much better than centralized authoritarians—and it may not stay democratic for long.
Similar difficulties arise when right-wing nationalists like Trump push for government stewardship of large sectors. See my 2024 piece, “The Case Against Nationalism,” co-authored with Alex Nowrasteh of the Cato Institute, for a fuller discussion. The striking overlap between Trump’s tactics and Sanders’s ideas illustrates how both socialist and nationalist currents can converge on fundamentally flawed policies. It’s a practical demonstration of what people sometimes call the Horseshoe Theory in action.
Supporters of Sanders who favor progressive aims should ask themselves whether they really want AI—and any substantial industry—under the control of a figure like Trump. He is not the first populist on the right to gain power, nor is he likely to be the last. Advocating powers that you would not want your opponents to wield is a precarious position.
The notion that the only alternative is a world dominated by a handful of billionaires is simply false. The AI market today features real competition. Competitors such as Claude, ChatGPT, Grok, Perplexity, and others vie for attention, each developed by different teams. New entrants continue to join the field. The owners of these companies understand they must satisfy consumer demand more effectively or at lower cost to succeed—provided they are not propped up by government subsidies or cronyism that state control tends to foster.
AI certainly carries risks, and there are valid arguments for restricting certain uses—especially in warfare and state surveillance. The prudent path is to curb dangerous applications rather than to pursue broad expropriation. Concentrating authority in the hands of politicians and their allies would likely increase those risks rather than mitigate them.
In conclusion, the proposal to seize a large portion of the AI sector is unconstitutional and poor policy by any standard. Its thrust aligns with the kinds of economic policy favored by Trump, rather than offering a sound alternative for innovation and growth.
