Anthropic's IPO structure leaves shareholders with little board control

Anthropic was founded in 2021 on the premise that humanity needs to keep control of AI and share in its long-term benefits. As the AI lab approaches its initial public offering, it has built one of Silicon Valley’s more elaborate governance structures — one that leaves ordinary stockholders who buy shares on public markets with little control over who sits on its board or how most major decisions get made.

Anthropic is organized as a public benefit company, a for-profit structure that lets its board weigh its stated safety mission alongside shareholder returns without facing shareholder lawsuits over decisions not driven purely by value. Current board members include chair and Anthropic president Daniela Amodei, her brother and CEO Dario Amodei, former Microsoft and GM finance chief Christopher Liddell, Netflix cofounder Reed Hastings, Spark Capital partner Yasmin Razavi and Novartis CEO Vas Narasimhan, with one seat still vacant.

A second layer of control sits with the Long-Term Benefit Trust, an independent panel holding a class of stock that can elect or remove a majority of the board’s “Class T” directors. The trust is chaired by global-health expert Neil Shah, a former GiveWell managing director, alongside Harvard physicist Christopher Stubbs, former Federal Reserve chair Ben Bernanke and national security expert Richard Fontaine; its single remaining vacancy will be filled by the trust’s own existing members. A third layer belongs to Anthropic’s seven founders, who jointly own a special stock class that hands a controlling bloc among them 50.1% of shareholder voting power and lets them elect three board directors — unless the founders split without a majority, in which case their shares instead vote in proportion with everyone else’s.

The board retains one check on the trust: if the Long-Term Benefit Trust nominates or removes a “Specified” director without alleging misconduct, Anthropic’s board can decline to go along and put the matter to a vote of the company’s employees — a provision that echoes OpenAI’s so-called “blip,” when its board briefly fired CEO Sam Altman only to reinstate him after employee loyalty forced a reversal. Founder control lapses automatically ten years after the IPO, or sooner if the founders choose to end it or their number falls below three; the trust remains active until common shareholders cast enough votes — at least 85% of them — to shut it down, or the board signs off on its own dissolution.

Why it matters

Until all of that falls away, power at Anthropic runs through its founders, a safety-focused trust and its own employees rather than through the capital markets — meaning investors buying into the IPO should expect limited influence over the AI lab’s direction for years.