Senator Bernie Sanders has proposed creating an American AI public wealth fund financed by the gains generated by artificial intelligence. His American AI Sovereign Wealth Fund Act would require a one-time 50 percent equity transfer from the nation’s leading AI firms into a federally managed investment fund. The government would hold voting shares and receive board representation, with the stated goal of ensuring that workers share in the benefits of technological progress as artificial intelligence transforms the economy.
The proposal differs fundamentally from the world’s best-known sovereign wealth and public pension funds. Norway’s Government Pension Fund Global was built from petroleum revenues. Canada’s pension fund invests mandatory worker and employer contributions. Sanders instead proposes creating public ownership through a transfer of private corporate equity. Whether such a proposal is legally or politically feasible in the United States is uncertain, but it raises a broader question with deep roots in American economic history: if government acquires substantial financial assets, how should they be governed?
Recent polling suggests that a majority of American workers support some form of AI public investment fund or dividend. Meanwhile, the United States faces mounting demographic pressure as the baby boom generation retires and Social Security’s long-run financing becomes increasingly strained. The rise of new wealth generated by either the AI tech boom or other sources creates a tempting source of new revenue to tax. Populist policymakers are likely to ignore incentives for continued innovation and will seek to expropriate some portion of those gains to help finance retirement security.
The key issue: effective, independent managers
History suggests that the central challenge is not accumulating public wealth. It is designing institutions capable of managing that wealth while limiting political interference and preserving the incentives that generate economic growth.
The first lesson comes from the creation of Social Security itself. As Carolyn Weaver documented in The Crisis in Social Security (1982), the original 1935 Social Security Act envisioned accumulating substantial payroll-tax reserves. Policymakers debated whether those reserves should eventually be invested in private stocks and corporate bonds, potentially making the federal government one of the nation’s largest investors.
Business organizations such as the US Chamber of Commerce and the National Association of Manufacturers strongly opposed that possibility. So did Republican Senator Arthur Vandenberg of Michigan. Although Vandenberg supported Social Security itself, he worried that a large publicly managed investment portfolio would concentrate excessive economic and political power in Washington. During Senate hearings, he pressed Arthur Altmeyer, the first chair of the Social Security Board, about how those reserves might ultimately be invested and what influence government ownership might create over American business.
Importantly, these concerns were not confined to conservatives. Many supporters of the New Deal also worried about placing enormous financial resources under political control. The resulting compromise restricted Social Security reserves to special Treasury securities rather than private corporate investments. Within only four years, Congress had largely abandoned the original reserve-fund model and shifted Social Security toward pay-as-you-go financing. The episode illustrates how quickly debates over public investment become debates over political power and institutional design.
The governance questions identified during the 1930s have never disappeared. Whenever government becomes a significant owner of financial assets, questions inevitably arise about who exercises voting rights, how investment decisions are made, and whether political objectives begin to influence commercial decisions.
Those concerns are especially relevant for artificial intelligence because Sanders’s proposal explicitly contemplates federal voting shares and board representation. But the underlying issues would exist even under a more passive investment structure. Public shareholders must still vote on corporate matters, and elected officials would inevitably face pressure to use those votes to pursue objectives beyond maximizing long-run returns.
Modern experience demonstrates that these governance problems can be managed, but only through institutions deliberately designed to separate investment decisions from day-to-day politics.
Canada’s Pension Plan Investment Board provides perhaps the clearest example. Created by Parliament in 1997, CPP Investments now manages more than C$700 billion in assets. Its directors are selected through an independent process designed to emphasize professional expertise rather than partisan affiliation. The board, not elected officials, appoints management, determines investment policy, and oversees compensation. The fund’s assets are legally separated from general government finances, and changing its governing legislation requires broad agreement among Canada’s provinces and federal government. Those institutional protections cannot eliminate political pressure, but they substantially reduce opportunities for short-term political intervention.
Norway illustrates both the strengths and limits of such arrangements. Its Government Pension Fund Global, now worth roughly $2 trillion, is widely regarded as the world’s best-governed sovereign wealth fund. Yet even Norway continues to face political controversy over ethical investment guidelines, exclusions of particular companies, and parliamentary debates about how the fund should exercise its ownership rights. Even highly professional governance structures remain vulnerable to political pressure. That reality would not have surprised participants in the Social Security debates of the 1930s.
Today’s AI proposals introduce an additional challenge that earlier sovereign wealth funds largely avoided. Norway’s wealth originated in oil. Canada’s assets originate in compulsory pension contributions. Artificial intelligence, by contrast, is not a natural resource. Its value depends on continued innovation by private firms operating in highly competitive global markets.
That distinction matters because policies affecting ownership also influence incentives. Firms and investors respond when governments target particular industries through taxation, regulation, or ownership. Such policies may reduce investment, discourage entrepreneurship, encourage activity to move elsewhere, or slow innovation. Those effects may be modest in some industries. They could prove considerably more important for artificial intelligence because AI is a general-purpose technology whose applications extend across nearly every sector of the economy.
The leading AI firms occupy central positions within a rapidly evolving innovation ecosystem. Their investments generate spillovers benefiting thousands of downstream firms and consumers. Policies that substantially alter incentives for founders, employees, investors, or future entrepreneurs therefore risk affecting not only the profitability of existing firms but also the pace of technological progress itself.
Don’t kill productivity
Economic security ultimately depends on productivity growth. A society with more retirees can maintain rising living standards only if workers and firms continue becoming more productive. If public policy unintentionally reduces the incentives that generate innovation, it may shrink the very source of wealth policymakers hope to share.
This consideration distinguishes AI from many earlier discussions of public investment. Governance affects not only how accumulated assets are managed but also the economic environment that determines whether those assets continue growing. Institutional design therefore must address both sides of the equation: protecting public wealth while preserving incentives for private innovation.
A final historical lesson concerns credible commitment. The architects of Social Security initially envisioned a substantial reserve fund financed through payroll-tax surpluses. Yet within only a few years, political pressures fundamentally altered that design. Their concern was not simply whether reserves would earn attractive returns. It was whether future governments could resist the temptation to redirect accumulated assets toward changing political priorities.
Any future American AI wealth fund would face similar pressures. Governments confronting recessions, fiscal crises, wars, or new spending priorities would inevitably face demands to use accumulated assets for immediate purposes rather than long-term retirement security. Durable institutional safeguards—including independent governance, transparent investment rules, statutory protection against fiscal diversion, and clear limits on political intervention—would therefore be essential.
Whether public ownership arises through natural-resource revenues, pension contributions, or AI generated wealth, history points to the same conclusion. The greatest challenge is not collecting wealth. It is creating institutions capable of managing that wealth over decades while preserving public confidence, limiting political interference, and maintaining the incentives that make economic growth possible.
The Social Security debates of the 1930s identified that challenge nearly a century ago. Canada’s pension system demonstrates how institutional design can reduce political influence. Norway shows that even exemplary governance structures remain subject to political pressures. As Americans increasingly debate how the gains from artificial intelligence should be shared, they would do well to remember that the success of any public wealth fund will depend less on how it is financed than on how it is governed.
