Anthropic's IPO governance leaves public shareholders with little say over its board

Anthropic, the frontier AI lab founded in 2021 to keep humanity in control of advanced AI, is heading toward an initial public offering. But according to TechCrunch, anyone buying its shares on the open market will own a piece of a company they cannot meaningfully steer: a layered governance setup concentrates power in an independent trust and a bloc of founders, leaving common shareholders with little say over most major decisions, including who sits on the board.

A public benefit company built for safety

Anthropic is incorporated as a public benefit corporation, a for-profit form that lets directors weigh a stated mission alongside profit. In Anthropic's case, that mission is AI safety, and the status gives the board legal cover to make safety decisions that are not driven entirely by shareholder value without facing lawsuits from unhappy investors.

The board is chaired by president Daniela Amodei and includes her brother, CEO Dario Amodei, along with Netflix co-founder Reed Hastings, former Microsoft and GM CFO Christopher Liddell, Novartis CEO Vas Narasimhan, and Spark Capital partner Yasmin Razavi. One seat is currently vacant.

The trust that picks most of the board

The most distinctive layer is the Long-Term Benefit Trust (LTBT), an independent panel created to keep the company focused on benefiting humanity. The trust holds a class of stock that lets it elect and remove a majority of the board, whose members are known as Class T directors.

The LTBT is chaired by Neil Shah, a global health expert and former managing director of GiveWell, the nonprofit that ranks charities. Its other members are former Federal Reserve chair Ben Bernanke, national security expert Richard Fontaine, and Harvard physicist Christopher Stubbs. One seat is open, and the sitting trustees will choose who fills it, making the trust effectively self-perpetuating.

Founders hold 50.1 percent of the vote

Anthropic's seven founders jointly own a special share class that gives a majority of them 50.1% of shareholder voting power, enough to elect three directors. That arrangement holds only while the founders vote together. If no majority exists among them, TechCrunch reports, their shares are instead voted in proportion with common stockholders, a crack in the founders' front that could hand public investors more influence than the structure nominally promises.

An employee escape valve

The design also builds in a check on the trust. When the LTBT nominates a designated "Specified" director among its four board picks, or removes that director without cause, the board's nominating committee can decline to approve the move and refer the decision to a vote of company employees.

TechCrunch likens this to the episode at OpenAI in which its board briefly fired CEO Sam Altman, only to back down and reinstate him under pressure from loyal employees, after which a new governance structure was put in place. Anthropic's setup formalizes that kind of employee power, giving staff a defined route to overrule the trust on board seats.

A slow sunset for founder control

The founders' grip does not last forever. Ten years after the IPO they can choose to end their control, or it lapses automatically once fewer than three of them remain. Even then, the LTBT continues until 85% of common shareholders vote to end it, or until it decides to step down with approval from the board. Only once both special share classes are gone, as TechCrunch dryly notes, hopefully not because of an AI-driven catastrophe, do common shareholders finally take control.

Why it matters

Anthropic is one of a small number of labs building frontier models, and whoever governs it shapes decisions with consequences well beyond its investor base. This structure is a concrete answer to the question of who restrains a frontier lab: not the market, and not public shareholders, but an unelected, self-filling trust alongside a founder voting bloc. That may be precisely the point, since insulation from quarterly pressure is arguably what safety-focused governance means in practice. But it also means IPO buyers will fund a company whose direction they cannot set, and it establishes a template other AI firms may copy as they go public. The mechanics are less predictable than they look, too: a split among the founders or an employee revolt could redistribute power in ways the formal structure only hints at.