Anthropic’s One-of-a-Kind Governance Framework Will Lock Public Investors Out of Almost All Decision-Making As It Nears IPO

Anthropic was founded in 2021 on a core mission: ensuring long-term, broad human control over advanced AI systems, so that the technology delivers widespread, shared benefit for all people. But as the leading AI frontier lab moves toward its long-awaited initial public offering, one large segment of the public is explicitly excluded from that model of shared control: every regular investor that buys its stock on open public markets.

Founder-friendly multi-class share structures that let leaders like Elon Musk and Mark Zuckerberg retain full control of their public companies while holding only a small fraction of total equity are hardly a new quirk of Silicon Valley. But Anthropic’s decades-long focus on AI safety has spurred an unusually baroque, multi-layered corporate governance design that leaves common shareholders with effectively no meaningful say over nearly all major company decisions — including who gets to serve on its board of directors.

---

The Four Interlocking Layers of Anthropic’s Special Governance Structure

1. Public Benefit Corporation Legal Shield

First, Anthropic is formally registered as a public benefit corporation: a for-profit entity legally permitted to prioritize a stated public mission alongside maximizing shareholder returns, with its explicit public mandate being advancing AI safety. This status gives the sitting board clear legal protection to make AI safety-focused decisions that do not prioritize short-term shareholder value, without facing costly lawsuits from disgruntled investors.

As of its latest pre-IPO filings, the board is chaired by Anthropic president Daniela Amodei, and counts CEO Dario Amodei (her brother), Netflix co-founder Reed Hastings, former Microsoft and General Motors CFO Christopher Liddell, Novartis CEO Vas Narasimhan, and Spark Capital partner Yasmin Razavi as sitting members, with one unfilled board vacancy.

2. Long-Term Benefit Trust (LTBT) Supermajority Board Power

A huge share of board control is allocated to an independent Long-Term Benefit Trust, a separate panel explicitly designed to keep the company anchored to its public mission of delivering benefits to all humanity. The trust holds a unique Class T share class that grants it the unilateral power to elect and remove a majority of Anthropic’s full board, known as Class T directors.

The LTBT is chaired by Neil Shah, a global health expert and former managing director of GiveWell, the non-profit that conducts independent, rigorous evaluations of global charities. Its other sitting members include former Federal Reserve chair Ben Bernanke, national security policy expert Richard Fontaine, and Harvard physicist Christopher Stubbs. There is currently one open vacancy on the LTBT, which will be filled by a vote of the existing trust members — no external input, let alone input from public shareholders, is required.

3. Founders’ Super-Voting Block

Anthropic’s seven original co-founders jointly own a separate special share class that gives a majority of the founding cohort 50.1% of all total shareholder voting power. This block lets them directly elect 3 members of the company’s board of directors. The only scenario in which their shares fall back to proportional, standard voting alongside common stock holders is if the founding group cannot reach a majority consensus among themselves on any given vote.

4. Formalized Employee Veto Power Over Board Appointments

The final layer is a cross-check that limits even the LTBT’s power: if the LTBT nominates a new "Specified" Class T director, or removes a sitting Class T director without formal documented cause (i.e. no proven misconduct), the board’s nominating committee is permitted to reject the move outright and send the final decision to a binding company-wide vote of Anthropic’s full employee base.

---

Unpredictable Dynamics and Long-Term Sunset Clauses

This unusually complex governance structure comes with highly untested, hard-to-predict outcomes. The tech industry already saw a prominent example of a similarly unorthodox AI governance system collapse unexpectedly in 2023, during OpenAI’s so-called "blip": the non-profit governing board briefly ousted CEO Sam Altman over concerns he was not being fully transparent with them, before a near-full company walkout threat in support of Altman forced the board to reverse its decision, reinstate Altman, and replace its entire governance structure.

Anthropic’s framework does away with that unstructured, ad-hoc employee power, and formalizes that level of worker input into its official rules. The structure creates notable edge cases: if the founding cohort loses its unified front, common shareholders could end up with far more de facto voting power than the written rules initially suggest, while rank-and-file employees could choose to side with the sitting operational board over the high-profile group of LTBT trustees to block appointments the trust pushes through.

None of the special control structures are permanent, however. Ten years after the company’s IPO, the founders can voluntarily choose to end their super-voting control, or that block will expire automatically if fewer than three of the original seven founders remain active at the company. Once the founders’ special control lapses, the LTBT will remain in power until 85% of all common shareholders vote to dissolve it, or the trustees vote to step down with formal approval from the full sitting board. Only when both special super-voting share classes no longer exist (and ideally, that transition is not triggered by the kind of catastrophic AI safety failure the entire governance system was built to prevent) will regular common public shareholders finally gain the standard full control that comes with owning stock in a public company.