Anthropic, a public benefit corporation (PBC), continues to offer food for thought on corporate governance and its mission to develop AI responsibly and for the long-term benefit of humanity. The frontier AI laboratory is preparing to enter public markets with a plan that would give founders and the Long-Term Benefit Trust, an independent body established to safeguard that mission, effective power to choose every director, leaving public investors with no power to elect board representatives independently. This would shield those entrusted with pursuing the company’s public benefit mission from electoral challenge by public investors, including those who may seek stronger safeguards in the public interest.

On September 24, Reuters, citing The Information, reported a proposal to give Anthropic’s seven co-founders 50.1 percent of votes on key shareholder matters. A prospectus reviewed by Reuters now explains the mechanism: A majority of the founders would direct a single Class F share through a Founder LLC. The filing invokes protection of Anthropic’s mission as the proposal’s purpose. The special rights would begin to sunset when only two or fewer founders or their successors remained.

The same report sets out the board split. Anthropic’s Long-Term Benefit Trust would elect four of seven directors. Class A and Class F shareholders would elect the other three, with the founders holding the controlling vote. The trust’s majority is a check on the founders, though neither set of seats would be separately elected by public shareholders.

As I argued in a recent post, Anthropic’s proposed antitrust waiver raised the question of who should determine acceptable AI risks across the industry. Its IPO plan raises that question as it relates to the company’s governance, where the ability to challenge those decisions depends partly on who chooses the board.

There are precedents for such electoral insulation. Meta preserves its founder’s voting majority; Palantir protects a collective founder bloc. Snap’s public shares carry no ordinary voting rights. Professors Zohar Goshen and Assaf Hamdani explain how control can preserve an entrepreneurial vision through disagreement with investors. Professors Lucian Bebchuk and Kobi Kastiel show why its benefits may fade while the costs of entrenchment rise. Founder guidance has a serious justification. Its scope and duration remain contested.

Anthropic adds a public benefit obligation. As a Delaware public benefit corporation, it must balance shareholders’ financial interests, the interests of those materially affected by its conduct, and its stated public purpose. It may lawfully issue shares with unequal votes. Professor Jens Dammann found dual-class shares in nine of nineteen listed PBCs in his 2024 study. The legal compatibility is clear. The question concerns the justification for concentrating control in a company so explicit about its moral mission.

Professor Brett McDonnell identifies protected control as a way to preserve a public mission against investors seeking higher returns. In AI, that could sustain costly safety measures or delay a release. The same arrangement can also insulate those interpreting the mission from shareholder challenge.

Professor Oliver Hart and Professor Luigi Zingales’s work on shareholder welfare exposes the cost of that insulation. Investors can care about corporate conduct and accept lower returns to prevent harm. Their votes can give those preferences institutional force. Anthropic’s shareholders might demand greater caution about AI or question a decision to deploy it. Protecting the company from shareholder pressure also limits those demands. The company’s public benefit mission should not require excluding the judgments of investors who may share that commitment.

A controlled company can reserve a board minority for public shareholders. At the New York Times Company, Class A shareholders separately elect 30 percent of directors, rounded upward. A mission guardian could retain a majority while public investors elected a minority. That would give investors a foothold in board elections, while leaving the wider question of public accountability open.

Professor Rebecca Henderson’s Moral Firms? places human flourishing at the center of corporate purpose. Professor Margaret O’Mara’s response asks whether purposes defined by executives and investors adequately reflect the societies their firms transform. Claude’s constitution endorses democratic institutions, oversight, and checks on concentrated power. Those commitments invite scrutiny of the company’s own allocation of authority.

Professor Emily Winston calls the institutional problem the separation of benefit and control: A PBC’s intended beneficiaries generally acquire no corresponding governance rights. Equal shareholder votes would not make investors representatives of the public. But restricting their voice removes one way to challenge judgments made in the public’s name.

OpenAI already illustrates the limits of formal oversight. Its board’s dismissal of CEO Sam Altman in 2023, followed by an employee revolt and his return, exposed the gap between the authority to remove an executive and the capacity to sustain that decision. Elon Musk’s claims concerning OpenAI’s charitable commitments were dismissed as untimely in May, leaving their substance unresolved. Its restructuring retained nonprofit control over a commercial PBC. Professor Moran Ofir and senior lecturer Ronit Levine-Schnur warn that formal nonprofit oversight can yield to commercial power. The question is who can enforce the mission when those running the business stray from it.

Anthropic’s draft S-1 prospectus places AI’s promise to society beside existential risk. Its governance therefore concerns power over the conditions of economic and political life. DARPA helped direct the early internet; industry now leads frontier model development. Where public leadership supplies few effective counterweights, company law helps determine who can set technological priorities, reorganize industries, and shape social life. Shareholder voice is one check, and public institutions must supply others. Montesquieu warned that “virtue itself has need of limits.” For Anthropic, the question is who can check those entrusted with its mission.

Marco Mari is a PhD candidate in business and social law at Bocconi University and a research fellow at NYU Law School’s Program in Corporate Law and Policy and the MIT Industrial Performance Center.