Etzioni on AI: Uncle Sam wants to participate in leading AI companies – what could go wrong?
Bernie Sanders and Donald Trump agree on almost nothing. But they agree that the public should have a stake in the AI industry.
The Vermont senator and the president disagree on both the structure and holding of public ownership, but the idea is being discussed at the highest levels of government. Even the later versions of OpenAI and Anthropic for the idea, although Anthropic is a tax instead of a pleasure. Let’s tune in.
The table below summarizes the initial proposals and shows how far apart they are, from voluntary outbreaks to outright capture. After participating in Intel, the president said he wanted “more cases like it.” Treasury paid $8.9 billion for 9.9% of Intel in August 2025; by the following spring the stake was worth about $36 billion, fueling interest in such deals. The Pentagon has already taken 15% of the rare miner. This is a pattern, not a one-off.
The argument for these proposals is a public finance argument, and a strong one. The science behind AI grew out of decades of federally funded research. Training data comes from the writing, code, and creativity of millions of people who are never solicited and never paid.
Sanders clearly states this principle: “When a public resource produces wealth, the public must share in that wealth.”
The purest versions cost the taxpayer nothing up front, because equity is donated rather than bought. That’s not the Intel model, which Washington bought for cash; it is the AI version that is now on the table, where the shares will be offered. When the bubble bursts, society gets nothing. If it agrees, the community has a piece. Unnecessary betting is a rarity in public finance.
| The source | A pole | The structure |
|---|---|---|
| Bernie Sanders | About 50% government position (reported figures vary) | Federal sovereign wealth fund; the government has voting rights; ~$1,000-per-person allocations |
| The Trump administration | Case by case statistics; 9.9% of Intel (now ~$36B) | Direct state ownership; it is included as a “windfall” for the taxpayer |
| OpenAI | ~5% of equity (~$42.6B) voluntarily contributed | A “Public Wealth Fund” made of Alaska’s; returns are distributed to citizens |
| Anthropic | There is no equality | Taxes on AI companies to fund employee support, possibly UBI |
Proposals from July 2026; Negotiations are still beginning and any version of the union will require an act of Congress.
However, there is a real risk of what happens when the government takes control of part of the industry it is supposed to control. Public participation in AI can be a boon or a pitfall, and the difference lives in the fine print.
Three things separate a dividend from a trap. The first is the size of the pole. The second is the wall between the government as the owner and the government as the referee, so the hand that keeps the budget has never written the safety rules. The third is hedging money: money reserved for employees is removed by technology, it does not go into the general fund. None of the three are compelling.
Here is a loose historical example. In 1998, 46 states settled with the tobacco industry for about $206 billion, paid off over 25 years. The states even depended on annual checks, which made them tacitly partners in the survival of the product they were supposed to fight against. And the money disappeared: today states spend only about three cents of every cigarette dollar on anti-smoking programs the fee was meant to fund.
A stake with no expiration date makes the government the owner of the industry it controls in perpetuity, and perpetuity is what turned tobacco land into tobacco. The answer is a fixed end date. The same statute that creates the stake must set the year it must expire. This is known as the sunset clause.
If Uncle Sam owns the stake, he must collect the dividend during the years of construction, then sell it on a fixed, published schedule until the position is liquidated. Ten or 15 years. Economists can choose a number. The deadline must be legislated from the beginning, so future Congresses cannot silently extend it.
Temporary co-ownership allows the community bank to look higher without leaving the referee with shares in the game entirely. Sanders and Trump, from opposite ends of the political spectrum, seized on the real grievance and reached for a permanent version of the solution, a version that is likely to be delayed. Of course, sunset stages aren’t carved in stone either.
Another challenge is that while Washington is in possession of the pieces of its AI masters, other topics follow – Beijing, Brussels, the Gulf – each takes its part, and the claim that American platforms respond that no government finds it difficult to implement. The dealer with the state on the cap table is not neutral. No wall and no expiration date solve this problem.
The stake also puts government in the business of picking winners. Owning a piece of OpenAI or Anthropic and Washington gains a financial interest in their business, and a reason to love them when they write the next law or sign the next contract. The startup is forced to compete with incumbents who are favored by offers. And in the fast-moving field of AI, the players are changing fast.
The economic challenges of AI are real and the appeal underlying these proposals is legitimate, but government ownership is the wrong solution. Conflicts of interest are real, precedents are bad, and it’s hard to imagine a referee with money in the game being neutral.
However, the pressure is real, too. Sanders, Trump, and the labs are all pushing versions of the same idea, and one of them can pass. If it does, the temporary version with guardrails beats the permanent one: price honestly, wall it off, direct money to damage, give a firm end date. None of this is a reason to take a pole. It’s the only thing that keeps a bad idea from getting worse.
