Anthropic S-1: The 7 Founders Who Control 50.1% of the Vote

Anthropic's Form S-1, filed with the SEC on October 26, 2026, is a long document about a fast-growing AI business. The most consequential sentence in it is about a single share of stock.

Immediately before the offering closes, Anthropic, PBC will issue one share of Class F common stock to Anthropic Founders, LLC, a vehicle owned by its seven co-founders. That one share carries 50.1% of the total voting power on matters put to shareholders generally, including the election of every director the company's Long-Term Benefit Trust does not appoint. It has no economic rights at all: no dividends, nothing in a liquidation beyond par value.

Inside the LLC, each founder has one vote and the majority decides. So the people who actually control Anthropic's shareholder vote are seven named individuals: Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Chris Olah.

The seven Anthropic co-founders in the Founder LLC

The filing names the seven and says they "as a group, are distinctly equipped to be stewards of our mission." All seven left OpenAI between 2020 and 2021 and founded Anthropic together in January 2021. Roles and share counts below come from the S-1's management, co-founder and beneficial-ownership sections.

Co-founder	Role at Anthropic	Class A shares (Sep 30, 2026)
Dario Amodei	CEO, director	240,637,980
Daniela Amodei	President, Chair of the board	241,408,940
Tom Brown	Chief Compute Officer	244,173,040
Jack Clark	Leads the Anthropic Institute	242,271,840
Jared Kaplan	Chief Science Officer	242,351,170
Sam McCandlish	Chief Architect	231,809,780
Chris Olah	Leads interpretability research	242,882,420
Total		1,685,535,170

Dario Amodei's figure excludes a further 4,621,120 shares held by donor-advised funds that he votes by irrevocable proxy.

Dario Amodei and Daniela Amodei

The two siblings sit at the top of both the company and its board. Dario Amodei, 43, has been CEO since founding. Before that he spent four and a half years at OpenAI, ending as Vice President of Research, after a stint as a senior research scientist at Google Brain and earlier work at Baidu's AI lab. He holds a Ph.D. in biophysics from Princeton. Daniela Amodei, 39, is President and Chairperson. At OpenAI she held VP roles in safety and policy and in people; before that she spent five years at Stripe in risk.

Tom Brown, Jared Kaplan and Sam McCandlish

These three are the research spine the S-1 credits with Anthropic's early credibility. Tom Brown is the first author of the GPT-3 paper, and the filing says he "led the engineering of GPT-3" and co-invented deep RLHF with Dario Amodei. He now runs compute, the largest cost line in the business. Jared Kaplan and Sam McCandlish are the first two authors of Scaling Laws for Neural Language Models, the 2020 paper the S-1 says was "critical in establishing" Anthropic's early relationships. Kaplan, a theoretical physicist, is Chief Science Officer and also helped pioneer Constitutional AI. McCandlish, also a physicist, previously served as Anthropic's CTO and is now Chief Architect.

Jack Clark and Chris Olah

Jack Clark came to AI from journalism, covering supercomputers and neural networks for The Register and BusinessWeek before becoming OpenAI's Policy Director. He writes the Import AI newsletter, leads the Anthropic Institute, and is one of four initial managers of the new Anthropic Public Benefit Fund. Chris Olah, a former Thiel Fellow who never finished a degree, led interpretability at OpenAI after Google Brain and is described in the filing as "a foundational figure in the field of mechanistic interpretability."

The eighth name: Ben Mann

One footnote does unusual work. It says Ben Mann "has been an integral part of Anthropic since its founding in January 2021" and "is often referred to as a co-founder." It then states plainly that he is not a member of the Founder LLC, not a Continuing Founder, and "will not participate in determining how our Class F common stock is voted." Everywhere else in the prospectus, "co-founders" means the seven.

Mann also worked on GPT-3 at OpenAI before leaving to start Anthropic. He now works in Anthropic Labs, the internal incubator behind Claude Code and the Model Context Protocol, where Anthropic's own announcement says Instagram co-founder Mike Krieger joined to build alongside him. The footnote reads as a deliberate line: Mann's founding role is acknowledged, but control was drawn around seven people.

The finding: 10% of the shares, at least 50.1% of the vote

The S-1 leaves the post-IPO ownership percentages blank until pricing, but it gives enough to compute the gap. The company says the 846,840,960 Class N shares it contributed to its Public Benefit Fund equal 5% of fully diluted shares as of June 30, 2026. That implies 16.94 billion fully diluted shares.

By that count, the seven founders together own about 10.0% of Anthropic (1.69 billion of 16.94 billion shares), while their LLC alone holds 50.1% of the shareholder vote. Method: sum of the seven founders' beneficially owned Class A shares as of September 30, 2026 (including options and RSUs vesting within 60 days), divided by the June 30, 2026 fully diluted count implied above. The two dates differ by one quarter, so treat the figure as approximate.

Source: Anthropic, PBC Form S-1, beneficial ownership table and Public Benefit Fund disclosure. Founders also vote their own Class A shares, so their real voting power is above 50.1%.

Two more things stand out in the numbers. The seven stakes are nearly identical, from 231.8 million (McCandlish) to 244.2 million (Brown), a spread of just 5.3%. Each founder owns roughly 1.4% of the company. One-person-one-vote inside the LLC mirrors an equity split that was close to equal from the start. And the 50.1% is fixed: the filing says it "will not be subject to dilution" from new share issuances, whether for fundraising, acquisitions or employee pay.

How the Founder LLC actually votes

The Founder LLC Agreement sets out the mechanics:

• Majority rule. The Class F share votes as a majority of Continuing Founders decide, one vote each.
• Deadlock goes to the public. If no majority forms, the share votes proportionately with the Class A votes actually cast, so a split founder group hands its weight to outside holders.
• Higher bars for some matters. On certain proposals the share votes "for" only if a supermajority of founders agrees; otherwise it votes "against."
• Personal shares stay personal. Each founder votes their own Class A stock however they like, independent of the LLC.

Founders can lose their seat in the LLC by resigning, through death or incapacity, by removal for cause (which includes specified competitive activity), by an approved leave of absence, or by selling down. The sell-down test is strict in one direction: a founder is out only if they keep less than 2% of the Class A shares they held at the IPO. Selling 98% still counts as staying in.

When the 50.1% ends

The structure has a sunset, but a slow one. If fewer than three Continuing Founders remain for 90 consecutive days, a 450-day Class F Transition Period starts. At the end, the company redeems the Class F share, and the employee-held Class E stock, for no consideration. Successors can be admitted when a founder dies or is incapacitated, but only until the tenth anniversary of the IPO. A supermajority of founders can also agree to retire the share early.

In practice, five of the seven would have to leave before the clock even starts, and then the founders keep control for roughly another 15 months.

Where the Long-Term Benefit Trust fits

The Founder LLC is one of two control blocks. The Long-Term Benefit Trust holds a single share of Class T stock that lets it elect four of the seven board seats, a majority. Class A and Class F holders, voting together, elect the other three, and because Class F is 50.1% of that vote, the founders decide those three. Under a stockholders' agreement, the Trust and the Founder LLC "will have the ability to elect all members of our board of directors."

The board today includes Netflix co-founder Reed Hastings as lead independent director, former Microsoft and GM CFO Christopher Liddell, Novartis CEO Vas Narasimhan, Spark Capital's Yasmin Razavi, and the two Amodeis. Anthropic will list on the Nasdaq Global Select Market as ANT and qualify as a "controlled company," which it says it will use only to keep a nominating committee that is not fully independent.

Other classes round it out. Class N shares, held mainly by strategic partners, have no vote. Class E, issued to employees other than the founders, votes only on narrow matters. The filing warns that the structure may keep Anthropic out of some stock indices and deter some institutions.

Why Anthropic built it this way

The business case for founder control is the S-1's own income statement. Revenue went from $41 million in 2023 to $386 million in 2024 and $4.59 billion in 2025, then $16.29 billion in the first half of 2026 alone, 14 times the same period a year earlier. The operating loss narrowed to $693 million in that half, while the headline net loss of $80.7 billion is almost entirely a $72.4 billion non-cash mark-up of convertible notes as the company's valuation rose.

The governance case is the one Anthropic makes in writing: as a public benefit corporation it must balance shareholder returns against "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." The risk factors say outright that the Trust and the founders "may vote in a way with which you disagree." Investors buying ANT are buying a share of the economics, and roughly half a vote, in a company whose direction seven people, not including Ben Mann, have agreed to steer together.

For a running account of the filing, see CNBC's S-1 coverage and The Next Web's breakdown of the Founder LLC.

Cover image: AI-generated illustration.