Who’s really in charge at Anthropic after the IPO?

An IPO usually means the public gets a vote, at least on paper. Anthropic's structure breaks that pattern. Board control is split between a Long-Term Benefit Trust, the company's seven founders through a special share class, and a backstop that can hand certain board disputes to employees. Common shareholders sit at the end of that line.

That matters because Anthropic is one of a handful of labs building frontier AI. Who can hire, fire, and overrule its leadership is not a corporate trivia question. It decides who steers a company that says its own technology carries catastrophic risk, and it decides whether outside capital can ever pull that steering wheel toward pure profit.

We think the instinct behind this design is defensible. A company that believes its models could become dangerous should not be one activist fund away from abandoning its safety commitments. The public benefit corporation status gives the board legal cover to weigh mission against returns, and the trust's power to elect a majority of directors is meant to keep that mission from being bought out. Given how fast this industry moves money, I understand the urge to bolt the doors.

But let's be honest about the trade. Buyers of this stock are being asked to fund the company while accepting that a small, self-selecting group decides most of what matters. The trust fills its own vacancies. The founders, voting as a majority bloc, hold just over half of shareholder voting power and pick three directors. That is a lot of trust placed in a lot of the same people, and accountability to the market is mostly deferred.

The part I'm watching most closely is the employee backstop. If the trust tries to seat or remove a specific director without cause, the board's nominating committee can refuse and send the question to a staff vote. On paper, that is a check on a panel of luminaries. In practice, it formalizes what we saw during OpenAI's leadership crisis, when employee loyalty to Sam Altman reversed a board decision within days. Writing that power into the charter could make Anthropic more stable, or it could make the workforce the real kingmaker in any future fight.

There is also a fragility here that the structure cannot fully hedge. Founder control depends on the founders agreeing. If they split, their shares vote with common stock, and suddenly the public matters more than anyone planned. Complex systems like this tend to behave predictably until the one day they are tested, and that day usually arrives during a crisis.

Our read: this is a serious attempt to keep a powerful company pointed at its stated mission, and it deserves credit for that. It is also a bet that the right people will stay in the right seats. If this model spreads across AI labs, we will end up with an industry where the most consequential technology companies are publicly traded but privately governed. That may be the safer path. It is not the more accountable one, and investors should price that in with clear eyes.

Anthropic was founded in 2021 with a stated goal of helping humanity keep control of AI and benefit from it. Its board currently includes Daniela Amodei as chair, CEO Dario Amodei, Reed Hastings, Christopher Liddell, Vas Narasimhan, and Yasmin Razavi, with one open seat. The trust is chaired by Neil Shah and includes Ben Bernanke, Richard Fontaine, and Christopher Stubbs, also with one vacancy.

The arrangement has an end date of sorts. Ten years after the IPO, founders can choose to give up their control, and it ends automatically if fewer than three remain. The trust then continues until 85% of common shareholders vote to end it or it steps down with board approval.