Senator Bernie Sanders wants the American public to own half of the country’s largest AI companies.
The senator said this week that he will soon introduce the American AI Sovereign Wealth Fund Act, which would create a public fund through a onetime 50% tax on the stock of major artificial intelligence companies. The public, in his telling, would get voting shares, board representation, and eventually a claim on the trillions of dollars that AI may generate.
While that might sound, at first, like another sweeping and unlikely bill from the Vermont Independent, its timing is interesting: A day later, President Donald Trump signed an executive order creating a 30-day federal review process for advanced AI models before public release.
The order is narrower than what some AI hawks have called for, and Sanders has already criticized its voluntary structure. But together, the two developments suggest that Washington is beginning to treat AI less like another software business and more like a strategic industry whose ownership, access, safety, and profits may be too important to leave entirely to private companies.
To be clear, a literal federal takeover of OpenAI, Anthropic, or Google DeepMind remains highly unlikely, at least outside some extraordinary crisis. But a softer version of nationalization—through prerelease review, procurement conditions, national-security contracting—is becoming easier to imagine.
Nationalization has layers
“There are different degrees of nationalization,” says Samuel Hammond, acting director of AI policy and chief economist at the center-right Foundation for American Innovation.
The bluntest version, he says, is outright government ownership: “where you literally buy the company.” That seems implausible to him, partly because the leading AI companies are now so expensive, and partly because the government would likely damage the very thing it was trying to control. Sanders’s proposal raises a narrower question: Should the public share in the wealth AI companies create?
Hammond is not endorsing a 50% stock tax, voting shares, or board seats for the government. But he does think there is a plausible case for what he calls “nationalizing part of the upside from powerful AI through something like a sovereign wealth fund.”
If a handful of private companies develop systems capable of automating huge portions of the economy, the financial gains could accrue to a remarkably small group of investors. A sovereign wealth fund, public equity stake, or some other public mechanism could spread that upside more broadly. “If these companies go to the moon,” Hammond says, “that means everyone gets a slice of that pie.”
Of course, the government does not need to buy AI companies to shape them. It can use procurement rules, military contracts, export controls, security designations, and public pressure from the White House. From Washington’s perspective, that’s much easier (and cheaper) than trying to somehow take control of a company. “This seems to be the much more likely pathway,” Hammond says.
The safety case for soft nationalization
The safety case for government involvement starts with a basic question: Who decides when an AI system is powerful enough to require public oversight?
That question has become harder to dismiss as frontier models move into cybersecurity and critical-infrastructure contexts. Anthropic’s limited release of Mythos, a model described as unusually capable at finding and exploiting software vulnerabilities, has become a fresh flash point in that debate.
The AI strategist Charles Jennings has been making the case for government oversight for years. In a 2023 Politico essay, he argued that the U.S. needed “to nationalize key parts of AI,” proposing a new “Humane AI Commission” modeled loosely on the Atomic Energy Commission (now the Nuclear Regulatory Commission), which controlled much of nuclear power in its early days. More recently, after Mythos renewed anxieties over frontier AI and national security, Jennings again called for government testing and approval of the most advanced systems. For him, the central problem is that AI is too powerful to remain governed mainly by the commercial incentives of a few frontier labs.
“We are not, as a society, coming to grips with this speeding train going down the track that’s exponentially getting faster and smarter and bigger and pulling more resources,” Jennings tells Fast Company.
He argues that AI is often misunderstood because people compare it to familiar technologies. “AI is a radical new technology, in the same way that the nuclear energy and an atomic bomb were radically new,” Jennings says.
The nuclear analogy is an admittedly imperfect one: AI isn’t a single weapon; it’s a general-purpose technology already spreading through seemingly every element of our lives. But Jennings uses the comparison to make a broader point: that some technologies are so powerful that society eventually decides they cannot be left entirely to private actors. “We need a new set of societal and global guidelines and safeguards to make sure that AI aligns with human values,” he says, “and we need to do it quickly.”
Jennings is not calling for the government to operate AI companies. Instead, he imagines something like a Food and Drug Administration-style model for frontier AI: Private companies keep building, owning, and commercializing their systems, but the most powerful models face expert review before release.
“The pharmaceutical companies own the IP, they do the innovation, they make the vaccines and the drugs,” Jennings says. “But they have to convince this group of experts that what they’re doing will be safe once it’s released in the wild.”
Trump’s new order doesn’t create that kind of approval regime, but it does move toward the premise that the government should have a look at the most powerful models before everyone else does. Jennings would argue that’s not a bad thing. “This stuff is just too powerful for it to be left in the hands of a few hypercompetitive CEOs,” he says.
The government problem
The risk is that government control can be dangerous, too. A federal AI agency could become slow, politicized, captured by industry, or weaponized by an administration hostile to civil liberties. A licensing regime meant to protect the public could entrench incumbents by making it harder for startups to compete. Military involvement could steer AI development toward surveillance and warfare.
That tension runs through the entire nationalization discourse. Few people want unaccountable corporate executives deciding the future of artificial intelligence by themselves. But handing that power to the federal government creates its own risks.
At some point, AI begins to resemble a utility more than a mere product. If intelligence becomes something people buy by the meter, like electricity or water, then governments may eventually ask utility-style questions: Who gets access? At what price? Under what reliability standards? With what obligations to the public? And with what limits on discrimination, manipulation, or harm?
For now, Sanders’s bill is unlikely to become law in its current form, and Trump’s model review order is far short of a licensing regime. But the two moves show how quickly the politics of AI are shifting. Hammond thinks the purest version of nationalization is “going nowhere,” and Jennings thinks some kind of government involvement is inevitable.
The government may never own the AI labs, but that’s not to say it won’t be involved in running them—one way or another.
