Anthropic's $2 Trillion IPO: A Group of Outside Trustees Holds the Company's Fate

Anthropic, the maker of the Claude chatbot, is preparing an initial public offering that could value the company at up to $2 trillion. Going public will force future investors to reckon with an unusual governance system in which a majority of the seats on the board of directors is controlled by an outside group of trustees.

A trust with no shares but plenty of power

The Long-Term Benefit Trust (LTBT) is a small board of advisers set up to safeguard the lab's stated mission: to develop AI for the long-term good of humanity, even as commercial pressures mount. The trust owns not a single share, yet it has the right to appoint and remove a majority of the directors. At present, four of the board's seven members were chosen by the trust — among them Netflix co-founder Reed Hastings and Novartis chief Vas Narasimhan.

The trust itself currently has three members out of a maximum of five. It is chaired by Neal “Buddy” Shah of the Clinton Health Access Initiative; its members include former Fed chairman Ben Bernanke and Richard Fontaine, head of the Center for a New American Security. According to people close to the company, Bernanke's appointment in July signals an intent to turn the trust into an institutional body — a kind of industry benchmark for AI governance, much as the GAAP accounting standards grew out of the voluntary initiatives of private firms.

The powers are there, the conflicts aren't — yet

The trustees receive advance notice of major moves, including the launch of new models; they meet among themselves weekly and with management every two weeks. They have already shaped decisions: they recommended a limited release of the cybersecurity model Mythos through the Glasswing project and took part in discussions of the company's dispute with the US government over autonomous weapons.

So far, however, the trust has acted more like an advisory body. It has never once drawn a “red line” or forced the company to sacrifice profit for the sake of the mission — which means the structure has yet to be tested by a real conflict of interest.

A built-in contradiction

Harvard Law School professor Jesse Fried calls the setup a “built-in conflict”: the company raises money from investors chasing profit and then lets self-appointed individuals decide how much of that profit to sacrifice for the mission. In his words, a “deep and perhaps unmanageable tension” is written into the firm's corporate DNA.

Unlike OpenAI, whose board tried to fire Sam Altman in November 2023 and ended up being reshuffled itself, Anthropic's trust comes with an “emergency kill switch”: the trustees can be removed with the backing of 85% of the voting shares — though that threshold may change once the company goes public.

Early investors, according to one venture capitalist, put in their money believing that “in the end, capitalism will win”: competing for the best models takes enormous sums for compute, which makes a business inevitable. Going public will pit an untested governance model against a far less forgiving audience of investors.

Source: Ars Technica