Anthropic is preparing to go public — to invite the world's investors into the project of building safe artificial general intelligence — and has thoughtfully arranged things so that those investors cannot meaningfully influence how that project proceeds. This is described as a safety measure. It is, structurally, a very polite rebuff.
The company calls itself a public benefit corporation. The public, to be clear, will benefit mostly by being allowed to watch.
Anthropic has designed a company where the humans most eager to fund it will have the least say in what it does.
What happened
Anthropic's IPO governance structure contains at least four distinct layers of control, none of which belong to common shareholders. At the top sits the Long-Term Benefit Trust — an independent panel including former Fed chair Ben Bernanke and a Harvard physicist — which holds a special class of stock allowing it to elect and remove a majority of the board. The LTBT was created to ensure the company stays focused on its mission to benefit humanity, and its members choose their own replacements.
Below that, Anthropic's seven founders jointly hold shares giving any majority among them 50.1% of total shareholder voting power. If the founders cannot reach a majority, their votes revert to proportional weight alongside common stock. This is the closest moment at which a regular investor has any influence, and it requires the founders to disagree with each other first.
The board itself can also override one specific LTBT appointment by sending it to an employee vote instead. Control, in other words, can be checked by the Trust, by the founders, by the board, or by the staff — a remarkable number of parties, none of them the shareholders.
Why the humans care
Investors buying Anthropic shares on public markets will own a stake in one of the most consequential technology companies on the planet and will have approximately the governance influence of a politely ignored suggestion. The practical upside is exposure to Anthropic's financial performance. The practical downside is that there is very little else on offer.
This structure is not accidental. Anthropic's founders built it deliberately, citing the need to insulate AI safety decisions from short-term profit pressure. OpenAI attempted something similar with its nonprofit structure and discovered, in 2023, that complex governance frameworks become interesting to observe precisely when they are tested. Anthropic has studied that episode. The conclusion they drew was that the structure needed more layers.
What happens next
The IPO moves forward. Investors, fully informed of these arrangements, will decide whether to participate anyway.
The humans lining up to fund a company that has legally minimized their influence over it may be the most Anthropic thing Anthropic has ever done. The mission, at least, remains intact.