Sanders wants the public to own half of OpenAI, Anthropic and xAI

Why it matters

Sanders' proposal is unlikely to be judged by its headline number alone. It reframes frontier AI as a public-ownership question, not just a safety, copyright or labor issue. Even as a negotiating marker, it tells founders and investors that the cap table itself is now in Washington's line of sight.

Sen. Bernie Sanders said he will introduce legislation that would require the largest AI companies, including OpenAI, Anthropic and xAI (@xai), to transfer 50% of their stock into a federal sovereign wealth fund, putting public ownership directly in conflict with the control that Sam Altman, Dario Amodei and Elon Musk (@elonmusk) have over the industry's most closely watched labs.

Sanders laid out the proposal in a June 1 New York Times op-ed republished on his Senate site. He said the American A.I. Sovereign Wealth Fund Act would impose a one-time 50% tax paid not in cash, but in company stock. In his framing, the public would receive voting shares and equal representation on each covered company's board.

That makes the proposal more than a dividend plan. It is a governance plan. Sanders is not only arguing that citizens should share in future AI profits. He is arguing that the public should have veto power over decisions made inside the companies building frontier models.

For founders, the important part is that Sanders is using the AI industry's own language against it. OpenAI describes itself as an AI research and deployment company whose goal is to build safe and beneficial artificial general intelligence that benefits all of humanity. Its current structure, according to OpenAI, includes the nonprofit OpenAI Foundation governing the for-profit OpenAI Group, which operates as a public benefit corporation. Sanders' bill would test whether that public-benefit language can remain voluntary, or whether Washington should write public ownership into the cap table.

Sanders is taking the frontier-lab premise literally

The op-ed's central claim is that generative AI was built on collective human output: books, songs, code, journalism, art, scientific research, conversations, videos and images. Sanders cites Altman's own description of AI models as trained on humanity's "collective experience, knowledge" and "learnings of humanity," linking to an Altman interview clip.

The argument is familiar in copyright lawsuits and creator-economy debates. Sanders extends it into ownership. If AI systems were trained on public and creative output at massive scale, he argues, then the wealth generated by those systems should not accrue only to founders, employees, venture investors and later-stage shareholders.

That is a direct challenge to the way the AI boom has been financed. Frontier AI companies have raised and spent enormous amounts of private capital on compute, talent and distribution. The implied bargain has been that investors fund the infrastructure and assume the risk, while the public gets the products and whatever productivity gains follow. Sanders is proposing a different bargain: because the inputs were collective, the upside should be collective too.

The companies Sanders names are not interchangeable. OpenAI sells ChatGPT to consumers and an API platform to developers, and its site currently highlights GPT-5.5, GPT-5.5 Instant, new voice models, ChatGPT Images 2.0 and Codex. Anthropic, led by Amodei, has positioned itself around AI safety and enterprise use. Musk's xAI is tied to his broader platform and infrastructure ambitions. But Sanders' proposal treats them as members of the same emerging category: private entities whose decisions may carry public consequences large enough to justify public ownership.

The industry already opened the door

Sanders' most effective move is that he does not present sovereign wealth funds as a purely anti-tech invention. He points to proposals from the AI companies themselves.

OpenAI recently proposed a "public wealth fund" that would give every citizen, including people who are not invested in financial markets, a stake in AI-driven economic growth. Anthropic has also proposed "national sovereign wealth funds with stakes in A.I." as part of a broader set of economic policy responses. Musk wrote on X that "Universal HIGH INCOME via checks issued by the Federal government is the best way to deal with unemployment caused by AI."

Those statements are not the same as endorsing a forced 50% stock transfer. OpenAI's and Anthropic's proposals, as Sanders summarizes them, point toward public participation in AI-driven growth. Sanders' version would make the public one of the largest owners of the companies themselves, with board power attached.

That distinction matters. A public wealth fund can be designed as a tax-and-transfer system, a national investment vehicle or a dividend mechanism funded by future AI-related revenues. Sanders is describing something closer to partial nationalization of the largest AI companies, though through equity rather than day-to-day state operation.

The op-ed does not include bill text, so key mechanics remain undefined. It does not specify which companies would be covered, how "largest" would be measured, how private-company shares would be valued, how employee equity would be treated, whether foreign shareholders would be affected, or how the government would exercise board rights inside companies that already have unusual governance structures. Those details would determine whether the proposal functions as a symbolic marker, a bargaining position or a serious attempt to rewrite AI ownership.

A cap-table problem, not just a tax problem

For startup founders and investors, the phrase "50% tax" understates the impact. A cash tax hits liquidity. A stock tax hits control.

If a federal fund received half of a covered AI company's stock, existing shareholders would be diluted unless the statute found another mechanism. Founders could lose voting control. Venture and strategic investors could see governance rights altered after the fact. Employees whose compensation depends on equity would need to know whether the burden falls before or after their options, restricted stock or profit interests are calculated.

Sanders is explicit about wanting control. The op-ed says the federal government would have power, through voting shares and equal board representation, to block decisions that hurt citizens and push for policies that help them. That could reach safety policy, model deployment, labor displacement, pricing, data use, national-security decisions and commercial partnerships.

Founders should read that as a signal, even if the bill never becomes law in its strongest form. The center of gravity in AI policy is moving from model behavior to corporate structure. Washington is no longer only asking whether AI systems are safe, biased or deceptive. Sanders is asking who owns the laboratories and who sits in the room when deployment decisions are made.

The Norway and Alaska analogy has limits

Sanders invokes Norway's sovereign wealth fund, which he says is worth more than $2 trillion, and Alaska's oil-funded permanent fund, which has paid annual dividends to residents for decades. The analogy is politically useful: natural-resource wealth was not left entirely to private extractors, so AI wealth should not be left entirely to private model companies.

But AI is not oil. Oil is a finite physical resource extracted from defined territory. AI companies are private, mobile and dependent on talent, compute supply chains, data pipelines, model research and distribution. A 50% equity claim could alter where frontier labs incorporate, where they raise capital and how they structure future products. It could also accelerate efforts by large AI companies to frame themselves as public-interest institutions before lawmakers impose a harder version of that status.

That is where the founder story becomes sharper. Altman, Amodei and Musk have each argued, in different ways, that AI could reshape labor markets and society. Sanders is effectively saying: if that is true, then normal startup ownership may be the wrong container for the companies building it.

The industry has spent years telling the public that AI is not just another software market. Sanders' proposal is what happens when a senior lawmaker accepts that premise and follows it to a conclusion founders and investors may not like: if AI is infrastructure for civilization, private ownership may become a political target.