An idea that until recently would have been seen as radical—that the public should co-own AI—now commands bipartisan consensus.
In June, Senator Bernie Sanders introduced the first AI tax in history, the American AI Sovereign Wealth Fund Act, which took up our AI equity tax framework. The bill proposes making the biggest AI companies owe the public half the equity in each company’s AI business, paid in newly issued shares.
In the weeks since, OpenAI’s Sam Altman and President Donald Trump have scrambled to compete, offering their own visions of an AI sovereign wealth fund, which, unlike a tax, would be voluntarily created through the companies’ partnership with the government. Specifically, the ChatGPT creator is reportedly considering giving the U.S. government a 5% stake in the company. We believe this approach is a mistake—and a tax is the solution.
The essential nature of a tax is that it’s mandatory, which is why a tax is the best way to secure the public’s standing as a co-owner of AI. Plus, the public supports an AI equity tax. Last month, a national poll found 69% of Americans in favor of requiring the largest AI companies to transfer half their stock into a public sovereign wealth fund.
The fight now is over the terms of public co-ownership: how much the public gets, who must deliver it, and whether delivery can be enforced. Altman and Trump are negotiating those terms with each other. Whether the public ever sees any benefit from the equity stake to be taken in its name is what hangs in the balance.
A deal between Altman and Trump would bind OpenAI to the current administration in the public’s name, at least symbolically. Yet the rights would belong to the administration, and the public would likely have no recourse if a later administration unwound the arrangement.
That brings us to an idea that until recently would also have been seen as radical: the federal government taking ownership of companies through a tax.
The Trump Administration has taken equity stakes in more than two dozen companies over the past year. The trend in those deals is that the rights belong to the administration, rather than being directly owed to citizens. And public dollars have to be handed over in exchange for the shares, putting the public on the hook.
Trump’s side of these deals follows a common pattern. His administration negotiates each stake company by company. In the case of Intel, government grants already owed were converted into an equity stake. In a mineral-rights deal, fresh taxpayer money was spent. And now, for something completely different, Trump is making early moves to accept AI stakes as donations. A government stake in AI companies “would be a beautiful thing,” the President says, and would ensure that “the American people can benefit from the success of AI.” Yet across Trump’s equity-acquisition deals, terms are established not by statute but through private negotiations, and the President or his agencies keep the rights.
Two deals illustrate how this works and why we should be wary. Trump publicly demanded Intel’s chief executive’s resignation. Only weeks later, the administration had its stake, a position that is now worth tens of billions. And as a condition for approval of the Nippon Steel acquisition, the administration took a golden share in U.S. Steel: “I, President Donald J. Trump, hold the Class G Preferred Stock (Golden Share) in U.S. Steel,” the President wrote in the Federal Register. The issue in both of these cases: one person, Trump, wields unprecedented power.
Altman argues a public stake is “the best way to share the upside of AI.” The company has reportedly proposed giving the administration a 5% stake, structured like the Alaska Permanent Fund and revocable whenever OpenAI chooses. The Financial Times reports the aim: clearing political obstacles by securing the administration’s financial buy-in. It’s a Trojan horse in which no one is fooled except the public.
We can guess where the public’s best interests could be treated as bargaining chips. OpenAI and Anthropic are preparing IPOs and likely need the administration’s goodwill. OpenAI, for example, is currently under pressure from the government to limit GPT-5.6’s release—a constraint the administration might be persuaded to drop in exchange for a donated equity stake. The administration’s record shows it accepts equity as consideration for looking the other way or relenting. Yet none of these dealings inherently benefit the American public.
In response, only Sanders has committed to actually issuing payouts to each American. With Sanders’ AI equity tax, there is no question mark about the public benefit of the tax. After all covered AI companies remit half their equity in newly issued shares, a public trust fund holds the shares, and the fund must pay out its returns to the public by statute. According to estimates from Sanders’ office, a 5% annual distribution would be about $1,045 per person per year. Since no public funds are used to buy shares and the bill specifically prohibits public bailouts of covered companies, the US public truly shares in the upside of AI.
The outcome of the midterms will likely determine whether Sanders’ AI equity tax advances, and thus the coming months may decide who owns America’s AI future. In the meantime, Trump and Altman may move ahead with their visions.
In our view, one design courts more corruption and wealth consolidation; the other ensures that, if AI ushers in any prosperity, it will be shared broadly, transparently, and with public accountability.