The AI goliath is here. OpenAI and Anthropic have each announced trillion-dollar fall IPOs. A central expectation underwriting those valuations is a vast transfer of wealth from labor to capital. AI companies’ models’ impressive capabilities will be deployed across numerous sectors to slash labor costs and payrolls, and AI companies’ shareholders will rake in profit. Despite this seismic shift in our political economy, as Senator Bernie Sanders recently observed, there has “not been one significant piece of legislation passed in Congress to address the impacts of AI.”
Meanwhile, progressives in Congress have also failed to utilize the full potential of tax law. The growing threat of AI offers a two-birds-with-one stone chance to change this, and in a recent article, we proposed just that: a framework for an AI company equity tax that both positions the public as owners with a voice inside the nation’s largest AI firms, and generates ample revenue in the process. On June 18, Senator Sanders put our idea into practice, introducing the American A.I. Sovereign Wealth Fund Act.
What this legislation promises is not only an important step in mitigating the risks of AI, but also a demonstration for legal academics and policymakers alike of how creative tax policy can serve the broader goals of the Left’s agenda. At the same time, it is worth contending with what a failure to seize this window of opportunity might portend—corrupt backroom deals that entrench rather than democratize AI’s growing threat.
The Promise of Taxing Equity
Taxes are a means to many ends. First, tax is a tool for fiscal capacity, allowing our public institutions to finance public programs like universal healthcare, affordable housing, and free college. Second, tax is a tool for redistribution, enabling greater levels of economic equality. And third, taxes can build public power so that democratic voice has greater control of our economy. By pursuing a forgotten but viable form of taxation—non-cash taxes—our AI equity tax proposal embraces all three possibilities.
To date, tax law has been constrained by a tacit premise that is rarely even discussed as a policy choice: taxes must be in the form of cash. While tech billionaires have long understood that the real action is in stock, and have hoarded their wealth accordingly, the public sector fiscal system has been shackled to cash. Cash taxes are ideal in many settings, but not all settings, and should be regarded as a choice. The Supreme Court affirmed just that in 1929, permitting Maryland to both assess and collect a tax in oyster shells to then be used for the general welfare of reseeding the beds of the Chesapeake for the benefit of future oyster farmers. Like cash, shells were considered “ordinary articles of commerce” by the Court and thus fair game under the taxing power. A tax collected in the form of corporate equity—as we propose—readily fits within this paradigm, offering modern policymakers an alternative means to taxing’s ends.
First, a tax on AI firms paid in stock, not cash, would raise significant revenue. In practice, this policy would assess an excise tax payable in newly issued shares, which would be contributed to a public trust akin to a sovereign wealth fund. Sanders’ proposed plan would claim a 50 percent ownership stake in the largest AI companies, worth an estimated $7 trillion. With such a large stake, publicly owned shares could be regularly sold off to generate revenue for the fisc without diluting the public fund’s influence over corporate governance. Further, the fund’s stock holdings could generate dividend income, provide leverage for general government borrowing, and furnish profitable derivatives out of the underlying assets.
Second, a tax on AI equity is highly redistributive, helping seed a future for all. Unlike the equity stakes purchased with the public’s money in more than twenty companies by President Trump during his second term, Sanders’ bill would not require any taxpayer funding for lucrative tech firms. Instead, the economic incidence of the tax would principally be borne by those with the greatest ability to pay—shareholders—diluting the value and influence of their holdings. The resulting fund would be obligated to make public distributions, offering a safety net to young workers trying to get a hold in a tightened labor market and to older generations getting laid off by AI. Both would have the added labor market bargaining power afforded by baseline financial stability.
The AI tax’s redistributive function also serves a corrective justice role, offering some compensation for all of the public’s labor and data that AI companies have scraped and used to train their models, often without consent. The public is entitled to reap these firms’ financial rewards—through co-ownership—because without training on troves of our data, the models are worthless.
Third, beyond the typical tax outcomes like revenue and redistribution, this proposal also offers democratic input into the economy. The public would have a voice, alongside private investors, in the vitally consequential choices of our country’s largest firms. Sanders’ plan guarantees board seats for the fund at the covered AI companies, which lets the public trust block decisions that harm Americans and push for ones that help them. Taken together, the taxation of AI illustrates how the Left could redesign the tax code on our own terms: using the taxing power to achieve fractional public ownership of systemically important firms, the benefits of which should be shared by all.
The Dangers of the Alternative
In this moment of opportunity, however, we must be wary of policy co-optation. Immediately following Sanders’ announcement of the legislation, OpenAI’s CEO met with the Senator to express his agreement in principle with some public ownership of AI firms, albeit not at Sanders’ proposed 50 percent. President Trump added that he also likes the idea, musing that he and Sanders are not far apart economically. Soon after, however, Trump and Altman indicated that the administration is working on a voluntary, rather than mandatory, mechanism for the public to own AI equity.
It is helpful to understand our proposal by illustrating what it is not. A voluntary, firm-by-firm contractual approach is not ownership, but simply another form of corruption, with AI companies bribing Trump to stall regulation. And Altman’s pitch of “free” equity for Americans does not pass the smell test. There is a trade going on, even if Americans never learn what Trump gives up that is rightly theirs.
In practice, the Trump-Altman approach would more likely yield non-voting stakes or contractual revenue-sharing rights negotiated behind closed doors, with passive financial upside in place of any democratic say. Further, a contractual stake could be unwound by this administration or the next, with no opportunity for the public to weigh in. Eschewing the transparency required by Sanders’ plan and its conflict of interest rules, the Trump-Altman plan transforms a broad public entitlement into a series of opaque private bargains.
Discretionary executive dealmaking also poses significant risk for the AI firms themselves. The recent incident in which Executive Branch directives forced Anthropic to abruptly pull its Fable model offline shows that relying on ad hoc government relationships in place of regulation leaves firms exposed to adverse, unpredictable, and opaque government actions.
As such, Sanders’ approach may be especially politically viable in appealing to the Altman ilk’s self-interest. A legislated tax within a comprehensive statutory framework offers the industry predictability and the rule of law, insulating companies from the transactional demands of supposedly voluntary government arrangements. Sanders’ bill, for example, vests the public’s stake in a proposed Independent Commission for Democratic AI, a bipartisan body modeled on the Federal Reserve to shield the fund from political pressure. Trump’s contractual approach offers no such insulation.
So, we find ourselves at a crossroads: follow a tax-based path to publicly held equity that establishes a durable, uniformly applied, and legally binding public stake, or remain paralyzed, allowing Altman and Trump’s corrupt alternative to fill the void.
***
AI has left us all in a precarious moment. But in precarity, there is opportunity for bold, creative thinking. And the potential of taxes paid in-kind, as we have proposed here, goes far beyond AI. For example, if all publicly traded firms listed on US exchanges had to remit 2 percent of all outstanding stock into a fund of this kind, the US could capitalize a trillion dollar reparations fund in less than a fiscal quarter! There is no shortage of wealth in the US, we just need new models for how to share it.
