Bernie Sanders just dropped the most aggressive public-benefit proposal for AI this century, and honestly? It's either going to be the most important policy speech of his career or the thing that makes him a permanent footnote. There's no in-between here.
The plan is simple enough to explain on a napkin: tax the stock of the biggest AI companies by 50%, park that money in a sovereign wealth fund, and hand dividends directly to every American. Sanders estimates the fund would be worth $7 trillion at full build-out, generating hundreds of billions annually. That's not pocket change. Every single American would receive more than $1,000 a year in 5% dividends — no means test, no application, just a check.
"The benefits cannot simply go to the handful of wealthy corporations," Sanders said. "They will be shared by the American people."
He's right, of course. The data that trained these models came from the public. The research that made them possible was funded by taxpayer dollars through DARPA, NSF, and university labs. The infrastructure they run on — broadband, power grids, semiconductor fabs — was built with public investment. And yet the profits flow to a handful of shareholders while workers get displaced and communities get hollowed out. Sanders isn't asking for charity from tech CEOs. He's demanding the return on an investment America made decades ago.
The question isn't whether the idea is sound. It's whether it can survive contact with reality.
How the Tax Actually Works
Here's where the mechanics get interesting, and where Sanders is being deliberately specific in a way that most politicians never are.
The legislation imposes a one-time 50% tax on the stock of any AI firm that does $200 million or more in annual AI sales. That threshold is low enough to catch the fast-rising startups, not just the incumbents. A new company that hits $200 million in AI revenue? Same tax. The law applies to existing firms and new entrants once they cross the line.
Sanders estimates the total fund would reach $7 trillion. At a 5% annual return, that generates roughly $350 billion per year — which he says would fund "direct payments to Americans and programs such as health care, education and housing." The per-person math checks out: $350 billion divided across 330 million Americans is about $1,060 each. Every year. Forever.
That's not a stimulus check. It's a permanent income floor funded by the returns on public ownership of AI infrastructure. Think about what that does to the political economy. You've got a program that can't be cut by Congress because it's self-sustaining. You've got a constituency of 330 million people who now have a direct financial stake in the AI industry's success. That changes the calculus on everything from regulation to trade policy.
But here's what makes this genuinely radical: the fund doesn't just hand out money. It gives Americans actual voting power over corporate decisions.
The Commission That Could Veto a Merger
This is the part that's going to keep AI CEOs up at night.
Sanders' legislation creates an Independent Commission for Democratic AI — seven members, bipartisan, nominated by the president and confirmed by the Senate. The commission holds voting shares in the taxed AI companies. That means it doesn't just collect dividends. It can block corporate decisions deemed harmful to the public.
"The public has got to have a significant seat at the table to make sure that terrible things do not happen to ordinary people, and that in fact, AI benefits ordinary people, not hurts them," Sanders told AP News.
Let that sink in. A government commission with actual board-level voting power could veto a merger, block a product launch, or force a change in how a model is trained. If OpenAI decides to deploy an AI system that the commission deems dangerous? They can stop it. If Anthropic wants to sell a product that exploits vulnerable populations? Veto power. This isn't oversight. It's ownership with teeth.
The commission structure is deliberately bipartisan — seven members, split between parties — to make it harder for either side to weaponize. But the mere existence of a government body with veto authority over private AI companies represents a fundamental shift in how we think about corporate governance. It's the difference between asking nicely and having a vote.
Sanders cast this as non-negotiable. "We think this is the best that we could do at the moment, and it's certainly a major, major, major step forward from giving unilateral and total power to a handful of multi-billionaires," he said.
The Corporate Split That Hits xAI Dead Center
If the tax and the commission were bad enough, there's another requirement in the bill that makes this truly disruptive: AI firms must split their non-AI business from their AI business.
This isn't a vague guideline. It's a structural break. The AI division has to stand on its own — separate infrastructure, separate board, separate finances. No more burying AI profits inside other business lines to avoid scrutiny.
This hits Elon Musk's xAI particularly hard. As The Hill reported, the legislation would require the company to untangle itself from its parent entities. xAI has already merged with X and then with SpaceX, creating a corporate structure that's deliberately opaque. And as the New York Times reported, Musk is reportedly plotting a potential SpaceX-Tesla mega-merger. Under Sanders' plan? That merger would trigger the split requirement, potentially forcing Musk to divest his AI holdings entirely.
The corporate breakup provision is the part of Sanders' plan that feels most like a return to the Trust-busting era. It's not about punishing success — it's about preventing concentration of power at a scale that no democratic institution can meaningfully regulate. When one entity controls the AI that runs your social feed, your car, your hospital system, and your defense infrastructure, you don't have a market. You have a monopoly on reality.
Sanders knows this. That's why the split requirement is in the bill.
The Industry's Panic: From Altman to Sacks
The reaction from the AI industry has been exactly what you'd expect: outrage, fear, and a lot of carefully worded statements about "constructive dialogue."
Sam Altman met with Sanders personally, and sources in the room told AP News that they remained "far apart" on how much stake in OpenAI the American public should have. Altman and Anthropic's Dario Amodei have shown some support for public benefits from AI — they've talked about revenue sharing and transparency — but their ideas are nowhere near as bold as Sanders'. Altman wants to be the benevolent dictator. Sanders wants to end the dictatorship.
Sanders didn't mince words about the alternatives being offered. "I think people like Sam Altman and Trump may be sympathetic to the public's growing concerns about AI displacing jobs or harming communities and are saying: 'OK, look, we're making zillions of dollars, so we're going to be nice guys and maybe we'll buy off the public. We will give 5 percent of our profits back into the government.' That's not what we're talking about. What we're talking about are two very different things," he said during the meeting.
He then cast AI firms that expect to transfer significantly less than 50% as "greedy." That's not a political statement. It's a challenge.
David Sacks, Trump's former AI czar, went even further. On the All In technology podcast, Sacks called Sanders' legislation "straight up confiscation of property" and said it sets "a terrible precedent." He told the Post he has "sympathy" for where Sanders is coming from but prefers voluntary public ownership ideas — like Trump's approach. Sacks is more supportive of CEOs choosing to give up a stake than of the government forcing them to.
The contrast is stark. Sanders wants structural change with legal force. The industry wants voluntary gestures that preserve the status quo.
The Political Math: Why This Won't Pass (But Still Matters)
Let's be honest about the odds. A Republican-controlled Congress that remains largely pro-AI industry is not going to pass this legislation without Trump's buy-in. And Trump has shown no interest in anything that looks like a tax increase, even one that technically comes from private companies rather than taxpayers.
Sanders likely doesn't expect his legislation to pass. He told AP News that he views it as a starting point for discussing how Americans should benefit from AI amid rising anti-AI sentiment nationwide. He knows the political headwinds. But he's also playing a longer game.
Anti-AI sentiment is growing across the country. Workers are displaced. Communities are hollowed out. The benefits of AI concentration flow to a tiny fraction of the population while the costs are distributed across everyone. Sanders' plan gives that frustration a concrete policy expression — $1,000 a year in your pocket, voting power over the companies reshaping your life.
Even if this bill never becomes law, it shifts the Overton window. It makes voluntary public ownership look timid by comparison. It forces every AI CEO to answer for why they think $7 trillion in public wealth should sit entirely in private hands. And it gives voters a clear alternative to the "trust us, we'll be nice" narrative that's been dominant for the past decade.
Sanders isn't trying to win this fight. He's trying to change the terms of the debate. And if history is any guide, that's how real change starts.