AI wealth carve-up is job best started right now

Amazon.com

Today 13:00

The author is a Reuters Breakingviews columnist. The opinions expressed are his own.
By Gabriel Rubin

    WASHINGTON, June 24 (Reuters Breakingviews) - Artificial intelligence is not only threatening to upend industries. The emerging technology is also re-energising debates about government ownership, regulation of critical industries and wealth redistribution. As the costs of building out AI and the possible economic and security ramifications of its development become clearer, politicians and executives are discussing the possibility of the U.S. government taking stakes in leading firms. Despite the many uncertainties, it's a task best started right away.
The discussion spans tech titans and elected officials of every stripe. Anthropic co-founder Dario Amodei and OpenAI boss Sam Altman, left-leaning Senator Bernie Sanders and President Donald Trump have all in recent weeks advanced ideas about how to deal with the economic, social and geopolitical fallout from AI, with ideas ranging from universal income programs to blunt job losses and quasi-nationalizations.
In many ways, these schemes are premature: the U.S. unemployment rate of 4.3% shows no evidence of AI-related job losses, while the “winners” of the boom are far from obvious. Nevertheless, waiting for a crisis before laying the legal groundwork for major changes to U.S. economic policy would be a mistake. Any attempt to transfer some of the burden of federal revenue from workers to AI ventures, for example, would be an epochal shift akin to last century's introduction of the federal income tax.
The competing proposals all treat AI as a discontinuity in American economic life different from anything that has come before. The similarities end there, however. Some focus on wealth redistribution, such as Sanders' plan to introduce legislation to create a public wealth fund with 50% stakes in AI companies. Trump wants a grand confab with the industry's top CEOs to discuss public ownership, possibly to defuse growing popular unease about data center construction and job losses.
A more conventional approach comes from Elizabeth Warren, who advocates corporate levies and special taxes on server facilities to fund social programs and enhanced unemployment and retraining benefits. The Democratic senator argues that states and the federal government should aggressively enforce existing laws and stop treating AI companies as untouchable.
The Trump administration has grown increasingly comfortable using federal money to fund private enterprise in return for ownership stakes. Since the president returned to the White House last year, Uncle Sam has taken stakes in U.S. Steel, various miners and chipmaker Intel INTC.O. It has done so using a variety of legal tools from the Defense Production Act, grants and loans from the federal Export-Import Bank, and a semiconductor manufacturing law passed under former President Joe Biden. Those authorities require either the consent of the companies or a compelling national security reason.
Any comprehensive quasi-nationalization drive would therefore have to start with new laws. Government agencies are forbidden from acquiring equity stakes in companies, especially a controlling one, if the firm is acting as an arm of the state. Legislation could establish the types of companies and activities that would be taxed or owned by the government, and where those proceeds would go, according to Peter Harrell, a former senior economic official in Biden's administration.
Yet even if Congress were to establish a sovereign wealth fund which held stakes in AI companies, it would need to constantly evolve in a way that similar state vehicles do not. Determining what an AI company is will only get trickier as the technology spreads and new businesses are started. It would be akin to identifying internet companies in the years following the dotcom boom, as the world wide web enabled everything from booking travel to online dating.
The fund's financial profile is also hard to discern. Any vehicle created today would have to take equity shares, as the likes of Anthropic and OpenAI are unlikely to make reliable taxable profits anytime soon. In Sanders’ plan, AI firms worth $200 billion or more would be subject to a 50% ownership stake by the United States. The senator estimates such a fund would be worth $7 trillion, though the legislation is light on calculations. Per the legislative text released last week, a 5% dividend yield would enable the fund to distribute $1,000 checks to taxpayers. Like all regulatory thresholds, however, this approach would encourage startups to stay smaller or shield their operations by arguing they are not primarily AI companies. The status of incumbent tech giants like Amazon.com AMZN.O or Alphabet GOOGL.O is equally fuzzy.
Any sovereign investment plan must account for these unknowns from the beginning. Alaska's $89 billion permanent fund balances these needs by reserving 25% of state oil revenue for future budgets, including the payment of an annual dividend that averages $1,000 - $2,000 per resident depending on fund performance. Its origins have echoes in the AI boom: following the lease of The Last Frontier's fields in the late 1960s, initial revenue was quickly exhausted by the state legislature, cutting Alaskans out of 1970s' booming oil sector profit until the fund was established in 1976.
In the shorter term, socializing potential future profit would mean subsidizing present risks. Equity stakes could leave U.S. taxpayers holding the bag for future cash calls as AI giants splurge on chips and computing infrastructure. Erstwhile White House AI czar David Sacks has pushed back hard on proposals for federal stakes, essentially accusing Anthropic of soliciting government support: “Signs you might be trying to get your frontier AI lab nationalized: You compare it to nukes…threaten half of white-collar jobs…warn recursive self-improvement could end humanity…then race ahead anyway,” he wrote on June 5.
A quasi-nationalization effort would raise broader questions, too. Would the U.S. government shy away from firmer regulation of AI companies if doing so undermined its investments? And how would other governments view American AI technology if the financial benefits were flowing in part to U.S. taxpayers?
All these caveats might persuade policymakers to simply rely on existing U.S. corporate and income tax systems to capture a part of the AI windfall, if and when it arrives. A fast-evolving industry doesn’t mean government should sit back and watch, however, even if managing a possible new industrial revolution seems a tough lift for the U.S. system. If politicians are going to intervene, it's best they start right away.
Follow Gabriel Rubin on Bluesky and LinkedIn.
CONTEXT NEWS
U.S. Senator Bernie Sanders said in June that he would introduce legislation for the U.S. government to take 50% stakes in artificial intelligence ventures, which would be held in a sovereign wealth fund.
U.S. President Donald Trump said on ​June 10 that he expects top artificial intelligence companies to agree to "giving ‌back" to the public, an apparent reference to a possible government stake in the firms.

(Editing by Peter Thal Larsen; Production by Pranav Kiran)
((For previous columns by the author, Reuters customers can click on RUBIN/gabriel.rubin@thomsonreuters.com))