Most legal responses to generative artificial intelligence ask familiar questions. Who may copy? Who may sue? Who should regulate? Who should pay? Jeremy Bearer-Friend and Sarah Polcz ask a more basic, and more corporate, question: who should own AI? Their answer: the public should own a piece.
Sharing the Algorithm sits comfortably in tax law, but corporate law scholars should read it because its ultimate subject is a corporate law mainstay: the ownership and control of enterprise. The article proposes a one-time tax on generative AI firms, paid not in cash but in equity. Instead of leaving the public to be only a consumer of AI products, a claimant in copyright litigation, a beneficiary of future tax revenues, or a regulator trying to catch up after the fact, Bearer-Friend and Polcz would make each of us a fractional owner of the firms that control generative AI.
The proposal is elegant because it connects the remedy to the source of the problem. Generative AI has been built through a vast extraction of human work, data, expression, likeness, and labor. Some of that extraction is already being litigated as copyright infringement. Some is showing up as labor displacement, some as discrimination embedded in algorithmic systems, and some as concentrated wealth and political power. Existing legal responses address these problems one at a time: licensing regimes might compensate some rights holders, excise taxes might raise revenue, and regulation might constrain some uses. None, however, matches the scale or character of the underlying social contribution that made generative AI possible in the first place. As just one example, the authors show that a solution focused on collective licensing faces a “missing authors” problem at grand scale: AI’s training corpus consists largely of everyday web content, not registered works. And, as they point out, a recent settlement gestures toward their alternative: Clearview AI compensated a plaintiff class not with cash but with a 23% equity stake in the company.
The authors’ key scholarly move is to shift the tax remittance from cash to equity. This matters because a cash tax can redistribute money, but it does not redistribute control. Equity can do both. If the public receives an ownership interest in AI firms, it can share in future gains while also obtaining governance tools: votes, information rights, board influence, and other elements of ownership. The point is not simply to make AI firms pay more, but to make the public less external to the technology that is reorganizing public life.
This is where Sharing the Algorithm becomes especially interesting for corporate law, and the authors are admirably concrete. Rather than invoking “public ownership” in the abstract, they work through four design questions: how to define the taxable category, what form of equity must be remitted, what rate should apply, and who should steward the resulting stake. On the form of equity, they consider several possibilities: requiring firms to remit the same kinds of securities already held by private investors; placing AI assets into a dedicated entity and remitting interests in that entity; or creating a specially designed instrument with specified rights to votes, board representation, profits, or information. On stewardship, they contrast direct distribution to citizens with a publicly managed fund modeled on public pension funds and sovereign wealth funds, with the Alaska Permanent Fund as one template. They are also careful about what the proposal is not: not receivership, not nationalization, and not a substitute for regulation, copyright remedies, antitrust, or ordinary taxation.
One of the harms the article targets deserves particular attention: labor displacement. Corporate law’s response to workforce concerns has often run through disclosure and board oversight, as reflected in the human capital management and human capital disclosure debates I have examined elsewhere.1 Those tools matter, but their reach is bounded by design. Bearer-Friend and Polcz propose something different: not better disclosure about the workers AI may displace, but a stake through which displaced workers, or the public on their behalf, may capture part of the returns of the displacing technology. By opening a channel for voice in corporate governance, the proposal moves the discussion closer to codetermination-style participation, an analogy the authors themselves invite, than to disclosure-based workforce governance alone.
Corporate law has long been comfortable with the idea that shareholders govern because they own and because they are the firm’s only residual claimants. That premise has always been asserted more forcefully than is warranted. As Summer Kim has shown, for example, residual claimant status need not turn only on the right to residual profits: it can also rest on firm-specific investment, exposure to risk, the wealth effects of corporate decisions, and monitoring capacity.2 Sharing the Algorithm pushes further, asking what follows when the inputs to value creation are far more widely dispersed than the cap table suggests. If generative AI is built from our writings, images, voices, conversations, workplaces, and social relations, then conventional ownership systematically understates who has contributed to the enterprise. The article’s power lies in making that mismatch visible and then offering an ownership-based response. Fractional public ownership accepts that AI may generate tremendous value, but insists that the value was made possible by public inputs and should therefore produce public claims.
Since the article’s publication, its core idea has migrated to the policy arena. OpenAI has proposed contributing a 5% equity stake to seed a public wealth fund that would pay dividends directly to citizens, modeled on the Alaska Permanent Fund. Anthropic’s leadership has discussed taxes on AI firms and long-term income support as responses to a world in which AI reduces demand for labor. And Senator Bernie Sanders’s American AI Sovereign Wealth Fund Act, introduced in June 2026, would impose a one-time 50% equity tax on AI companies with annual sales above $200 million and place the resulting shares in a publicly managed fund. Both sets of proposals concede the core intuition developed by Bearer-Friend and Polcz: the economic upside of AI should not flow only to existing shareholders.
If any of these proposals advance, hard questions follow, and many belong to corporate law. Because the most consequential AI developers are privately-held unicorns, often with idiosyncratic structures such as nonprofit status, capped-profit interests, and benefit corporation charters, the remitted equity may carry no market price, no exit, and thin information rights. The authors also concede that most shareholders are passive; whether a politically accountable public fund would be an effective monitor or a conflicted one remains unresolved. Bearer-Friend and Polcz leave these issues for another day, and this is less a flaw than an invitation.
Whether one ultimately favors an AI equity tax or doubts its constitutional and political prospects, Sharing the Algorithm reorients law’s encounter with AI. It invites corporate law scholars to see tax not only as a revenue device, and corporate equity not only as a financing instrument, but as a way of allocating voice over the future.
- See George S. Georgiev, The Human Capital Management Movement in U.S. Corporate Law, 95 Tul. L. Rev. 639 (2021); George S. Georgiev, Human Capital Disclosure & Corporate Governance: The New Evidence, 46 Cardozo L. Rev. 485 (2024).
- See Sung Eun (Summer) Kim, Dynamic Corporate Residual Claimants: A Multicriteria Assessment, 25 Chap. L. Rev. 67 (2021).
