AI People Graph · Lab 1: Anthropic · Updated 18 NOVEMBER 2026
Seven people left OpenAI and founded Anthropic in January 2021. This is who they are, how they were connected before, and how their shares and votes are wired into the company that has now filed to go public.
In one paragraph: Anthropic’s seven co-founders all came from OpenAI: Dario Amodei (CEO), Daniela Amodei (President and board chair; the two are siblings), Jared Kaplan, Jack Clark, Tom Brown, Sam McCandlish and Chris Olah. Before Anthropic, Dario Amodei had already written papers with 5 of the other 6. Anthropic’s S-1/A No. 3, filed on 17 November 2026, gives each of them between 240.6 million and 244.2 million Class A shares before the IPO, 3.6% to 3.7% of the class, and each sells 3,378,378 shares in the offering. Their real lever is one share: Class F, held by a Founder LLC in which each of the seven has one vote, carries 50.1% of the votes on stockholder matters, so with their own shares they control 61.3%. The Long-Term Benefit Trust, holder of the single Class T share, elects 4 of the 7 directors.
The graph
Hover or tap a person or an organisation to light up its connections; tap again (or click) to open the profile. Each line is one sourced connection, coloured by type; switch types on and off with the buttons. Lines between co-founders are their first paper together before 2021; turn on “Papers since 2021” to add the papers written at Anthropic.
Tap or hover a node
7 people, 14 organisations, 62 sourced connections. Every line has its source in the data files behind this page.
All connections as a list
- Anthropic: co-founded Anthropic, January 2021
- Baidu Research: Baidu Research, Silicon Valley AI Lab: First-listed author of Deep Speech 2 under that lab's name
- Google (Brain): Google Brain: Senior Research Scientist
- OpenAI: OpenAI: Vice President of Research, among other positions
- Daniela Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jared Kaplan: first paper together (in our set): An Empirical Model of Large-Batch Training (2018)
- Jack Clark: first paper together (in our set): The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation (2018)
- Tom Brown: first paper together (in our set): Deep reinforcement learning from human preferences (2017)
- Sam McCandlish: first paper together (in our set): An Empirical Model of Large-Batch Training (2018)
- Chris Olah: first paper together (in our set): Concrete Problems in AI Safety (2016)
- Daniela Amodei: siblings (S-1: 'Family Relationships')
- Anthropic board: director; elected by Classes A and F after the IPO
- Public Benefit Fund: initial manager of the Public Benefit Fund (the board decides its grants until the Fund's governance is approved)
- Donor-advised funds: proxyholder: votes certain DAF-held shares (4,621,120 in the table)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
- Anthropic: co-founded Anthropic, January 2021
- Stripe: Stripe: Risk Manager, among other positions
- OpenAI: OpenAI: VP of Safety and Policy, VP of People, Engineering Manager
- Dario Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jared Kaplan: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jack Clark: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Tom Brown: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Sam McCandlish: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Chris Olah: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Dario Amodei: siblings (S-1: 'Family Relationships')
- Anthropic board: Chairperson; elected by Classes A and F after the IPO
- Public Benefit Fund: initial manager of the Public Benefit Fund (the board decides its grants until the Fund's governance is approved)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
- Anthropic: co-founded Anthropic, January 2021
- Johns Hopkins: Johns Hopkins University: Associate professor of physics; taught the foundations of deep learning
- OpenAI: OpenAI: Researcher (the 2018 batch-size paper lists him at Johns Hopkins and OpenAI)
- Dario Amodei: first paper together (in our set): An Empirical Model of Large-Batch Training (2018)
- Daniela Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jack Clark: first paper together (in our set): Language Models are Few-Shot Learners (2020)
- Tom Brown: first paper together (in our set): Scaling Laws for Neural Language Models (2020)
- Sam McCandlish: first paper together (in our set): An Empirical Model of Large-Batch Training (2018)
- Chris Olah: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
- Anthropic: co-founded Anthropic, January 2021
- OpenAI: OpenAI: Policy Director: engagement with governments and regulators
- Dario Amodei: first paper together (in our set): The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation (2018)
- Daniela Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jared Kaplan: first paper together (in our set): Language Models are Few-Shot Learners (2020)
- Tom Brown: first paper together (in our set): Language Models are Few-Shot Learners (2020)
- Sam McCandlish: first paper together (in our set): Language Models are Few-Shot Learners (2020)
- Chris Olah: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Public Benefit Fund: initial manager of the Public Benefit Fund (the board decides its grants until the Fund's governance is approved)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
- Anthropic: co-founded Anthropic, January 2021
- Google (Brain): Google: Adversarial Patch lists him with a google.com address
- OpenAI: OpenAI: Led the engineering of GPT-3
- Dario Amodei: first paper together (in our set): Deep reinforcement learning from human preferences (2017)
- Daniela Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jared Kaplan: first paper together (in our set): Scaling Laws for Neural Language Models (2020)
- Jack Clark: first paper together (in our set): Language Models are Few-Shot Learners (2020)
- Sam McCandlish: first paper together (in our set): Scaling Laws for Neural Language Models (2020)
- Chris Olah: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
- Anthropic: co-founded Anthropic, January 2021
- OpenAI: OpenAI: Built and led a team on the science and scaling of machine learning
- Dario Amodei: first paper together (in our set): An Empirical Model of Large-Batch Training (2018)
- Daniela Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jared Kaplan: first paper together (in our set): An Empirical Model of Large-Batch Training (2018)
- Jack Clark: first paper together (in our set): Language Models are Few-Shot Learners (2020)
- Tom Brown: first paper together (in our set): Scaling Laws for Neural Language Models (2020)
- Chris Olah: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
- Anthropic: co-founded Anthropic, January 2021
- Google (Brain): Google Brain: Research scientist; Concrete Problems (2016) and Feature Visualization (2017) list him at Google Brain
- Distill: co-founded Distill
- OpenAI: OpenAI: Led interpretability research (Zoom In, March 2020, lists him at OpenAI)
- Dario Amodei: first paper together (in our set): Concrete Problems in AI Safety (2016)
- Daniela Amodei: first paper together (in our set): Predictability and Surprise in Large Generative Models (2022)
- Jared Kaplan: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Jack Clark: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Tom Brown: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Sam McCandlish: first paper together (in our set): A Mathematical Framework for Transformer Circuits (2021)
- Founder LLC: member of the Founder LLC; one vote on how Class F votes
How they connect
One lab, one exit
All seven worked at OpenAI before co-founding Anthropic in January 2021. Dario Amodei was there from July 2016 to December 2020 (Vice President of Research among other roles), Daniela Amodei from October 2018 to December 2020. Tom Brown led GPT-3’s engineering; Jack Clark was Policy Director; Chris Olah led interpretability.
Dario Amodei is the hub of the paper trail
Before 2021 he had co-written with Chris Olah (Concrete Problems in AI Safety, 2016), Tom Brown (learning from human preferences, 2017), Jack Clark (The Malicious Use of AI, 2018), and Jared Kaplan and Sam McCandlish (large-batch training, 2018). Daniela Amodei’s first paper with the others is the 2022 one all seven signed.
The scaling-laws core
Kaplan, McCandlish, Brown and Dario Amodei are 4 of the 10 authors of Scaling Laws for Neural Language Models (2020); with Clark, 5 of the 31 authors of GPT-3 (Language Models are Few-Shot Learners, 2020). Kaplan and McCandlish are physics PhDs (Harvard, Stanford); Dario Amodei’s Ph.D. is in biophysics (Princeton).
Google before OpenAI
Dario Amodei (Senior Research Scientist, October 2015 to June 2016) and Chris Olah were at Google Brain, and signed Concrete Problems in AI Safety under it in June 2016. Tom Brown’s 2017 Adversarial Patch lists a google.com address. Earlier, Dario Amodei is the first-listed author of Baidu’s Deep Speech 2 (2015).
11 of the 21 co-founder pairs had a paper together before 2021; by 2022 all 21 had. Sources for each line are in the graph’s tooltips and in the data files listed at the end.
How the shares and the votes move
Read each row left to right. The filing separates owning Anthropic (economics) from deciding for it (votes and board seats); the co-founders hold a modest slice of the first and, through one share, a majority of the second.

The Founder LLC, in six rules
- The seven are its only members at the start, and it holds the single Class F share. source
- Each Continuing Founder has one vote on how Class F is cast; a majority decides. Deadlock: Class F votes in proportion to the Class A votes cast. source
- A co-founder drops out on resignation, death or incapacity, removal for cause (including specified competitive activities), or if they and their related holders fall below 2% of the Class A they held at the IPO. source
- Successor members can be admitted after a death or incapacity until the tenth anniversary of the IPO. source
- Sunset: once there have been fewer than three Continuing Founders for 90 days in a row, a 450-day transition starts; at its end Class F (and Class E) is redeemed for no consideration. source
- Anthropic pays the LLC’s legal fees for setting this up, expected at about $5,000,000. source
Benjamin Mann, often called a co-founder, is not a member of the Founder LLC; the S-1 counts seven co-founders. source

The IPO sale
Each of the seven sells 3,378,378 shares; together 23,648,646, which is every share sold by selling stockholders on the cover (our arithmetic: 7 × 3,378,378). At the filed range of $70.00 to $78.00 that is $236.5M to $263.5M each and $1.7B to $1.8B for the seven, before underwriting discounts (our arithmetic). Anthropic receives none of it. source
Earlier sales: two tender offers
June 2025: Anthropic bought back 27,942,000 shares at $5.61 for $156,716,898; Dario and Daniela Amodei sold into it. April 2026: a syndicate of stockholders bought 179,883,540 shares at $25.9136 for $4,661,437,298; Dario and Daniela Amodei, CFO Krishna Rao and Holden Karnofsky (Daniela Amodei’s spouse, an Anthropic employee) sold into it. The filing gives totals, not each seller’s shares, and its related-person section covers directors, executive officers and 5% holders, so it says nothing either way about the other five co-founders. source
Pay

Pledges, proxies and the Trust
80% pledges. Each co-founder has pledged to dedicate 80% of their personal equity in Anthropic to philanthropic causes. The S-1 says these are “voluntary, revocable expressions” of intent, not legally binding, and that they cover Class A only: as the co-founders give, their economic stake falls while their Founder LLC votes stay the same. source
Donor-advised funds. Daffy Charitable Fund and Dechomai Asset Trust can receive Anthropic shares from the company or from stockholders; under irrevocable proxy agreements of 6 October 2025 and 20 December 2024, Dario Amodei votes their shares, except those given by him or his family. source
The Trust. The Long-Term Benefit Trust has five trustee seats, one vacant: Neil “Buddy” Shah (chair since June 2023), Richard Fontaine (since April 2025), Ben Bernanke (since June 2026) and Christopher Stubbs (since October 2026). Trustees may not buy or hold Anthropic securities; Anthropic paid the Trust $8,425,213 in the nine months to 30 September 2026. source
Amazon and Google. After the IPO they hold 14.7% and 13.7% of Anthropic’s capital stock, all in non-voting Class N. The Anthropic Public Benefit Fund holds 846,840,960 Class N shares, 5% of fully diluted shares at 30 June 2026, and the board may add up to 2% more each time the share price rises 50% above the prior milestone. source
The seven
In the order the S-1 lists them. Figures from the beneficial ownership table of the S-1/A No. 3 (as of 30 September 2026); careers from the S-1 and the papers’ own author lines; quotes only from the person’s own essays and pages.
Dario Amodei
Chief Executive Officer, co-founder and director, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
- Director
- Proxyholder for 2 donor-advised funds
- Public Benefit Fund initial manager
Career
- 2015 (paper)Baidu Research, Silicon Valley AI Lab: First-listed author of Deep Speech 2 under that lab's name source
- Oct 2015 – Jun 2016Google Brain: Senior Research Scientist source
- Jul 2016 – Dec 2020OpenAI: Vice President of Research, among other positions source
- Jan 2021 –Anthropic: Co-founder, Chief Executive Officer, director
- EducationPh.D. in Biophysics, Princeton; B.S. in Physics, Stanford; postdoctoral fellowship, Stanford University School of Medicine source
Key papers
- Deep Speech 2: End-to-End Speech Recognition in English and Mandarin 2015 · arXiv · end-to-end speech recognition, Baidu Research Silicon Valley AI Lab
- Concrete Problems in AI Safety 2016 · arXiv · five research problems in AI accident risk: side effects, reward hacking, scalable supervision and more
- Deep reinforcement learning from human preferences 2017 · arXiv · training agents from human preference comparisons
- Scaling Laws for Neural Language Models 2020 · arXiv · the neural scaling laws: loss falls as a power law in model size, data and compute
- Language Models are Few-Shot Learners 2020 · arXiv · the GPT-3 paper: few-shot learning at scale
Papers with the other co-founders before Anthropic: Chris Olah (2016); Tom Brown (2017); Jared Kaplan (2018); Jack Clark (2018); Sam McCandlish (2018). The year is their first paper together in our set.
His own site says he co-invented reinforcement learning from human feedback and led the development of GPT-2 and GPT-3 at OpenAI. source
Stake per the S-1. 240,637,980 shares beneficially owned at 30 September 2026: 237,813,080 held of record by him or by trusts he controls, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 237,259,602. S-1 table footnote
Votes he casts for others. He holds irrevocable proxies over Class A shares held by two donor-advised funds, Daffy Charitable Fund and Dechomai Asset Trust: 4,621,120 shares in the ownership table, on top of his own. Shares that he or his family give to those funds do not count. source
Pay (2025, as a named executive officer). $17,958,298 in total: salary $654,167, stock awards $9,264,932 and option awards $4,028,558 at grant-date fair numbers, and $4,010,641 of other compensation, of which $2,745,874 was personal security. His base salary is $1,380,000 since 1 July 2026. source
Earlier sales. He sold shares in the company’s June 2025 tender offer at $5.61 a share and in the April 2026 tender at $25.9136; the filing does not give his number of shares in either. source
Public Benefit Fund. The S-1 names four initial managers of the Anthropic Public Benefit Fund: Dario Amodei, Daniela Amodei, Jack Clark and LTBT chair Neil “Buddy” Shah. Until the board approves the Fund’s own governance, the board decides its grants. The Fund holds 846,840,960 Class N shares, 5% of Anthropic’s fully diluted shares at 30 June 2026. source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
“I think that most people are underestimating just how radical the upside of AI could be, just as I think most people are underestimating how bad the risks could be.”Dario Amodei, Machines of Loving Grace (October 2024)
“We can’t stop the bus, but we can steer it.”Dario Amodei, The Urgency of Interpretability (April 2025)
On FourWeekMBA
- 18 Nov 2026How Anthropic’s Public Benefit Duty Works: The Trust, Class F and the $2M Suit
- 18 Nov 2026Son Calls Missing Anthropic a Mistake After $64.6B on OpenAI
- 18 Nov 2026WSJ Ties Anthropic Stakes to SPVs; Filing Lists 7 Founders
- 17 Nov 2026Anthropic Said to Price IPO Above $80; Filing Count: $1.05T+
- 17 Nov 2026Amazon Gets 33.9% of Anthropic’s Class N at IPO, Not 48.2%
- 17 Nov 2026OpenAI Director’s 1-Year AI Takeover Risk: 10%, vs 22% Ever
Newest 6 of 129 pages that mention Dario Amodei.
Daniela Amodei
President, co-founder and Chairperson of the board, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
- Chairperson of the board
- Public Benefit Fund initial manager
Career
- May 2013 – Jul 2018Stripe: Risk Manager, among other positions source
- Oct 2018 – Dec 2020OpenAI: VP of Safety and Policy, VP of People, Engineering Manager source
- Jan 2021 –Anthropic: Co-founder, President, Chairperson of the board
- EducationB.A. in English Literature, University of California, Santa Cruz source
Key papers
- Predictability and Surprise in Large Generative Models 2022 · arXiv · the one paper in our set all seven co-founders signed
- Measuring Progress on Scalable Oversight for Large Language Models 2022 · arXiv · measuring progress on scalable oversight
- Discovering Language Model Behaviors with Model-Written Evaluations 2022 · arXiv · discovering model behaviours with model-written evaluations
Papers with the other co-founders: none before 2021 in our set; the first is Predictability and Surprise in Large Generative Models (2022), the one paper all seven signed.
Stake per the S-1. 241,408,940 shares beneficially owned at 30 September 2026: 238,319,540 held of record by her or by trusts she controls, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 238,030,562. S-1 table footnote
Pay (2025, as a named executive officer). $16,408,998 in total: the same salary, stock and option awards as her brother, plus $2,461,341 of other compensation, including $1,293,629 for personal protection, home security and transportation. Base salary $1,380,000 since 1 July 2026. source
Earlier sales. She sold shares in the June 2025 ($5.61) and April 2026 ($25.9136) tender offers; the filing does not give her number of shares. source
Related-person disclosure. The S-1 names her spouse, Holden Karnofsky, as an Anthropic employee since January 2025 who also sold in the April 2026 tender; her beneficial ownership includes 130,550 options and 133,950 RSUs held by him. source
Public Benefit Fund. The S-1 names four initial managers of the Anthropic Public Benefit Fund: Dario Amodei, Daniela Amodei, Jack Clark and LTBT chair Neil “Buddy” Shah. Until the board approves the Fund’s own governance, the board decides its grants. The Fund holds 846,840,960 Class N shares, 5% of Anthropic’s fully diluted shares at 30 June 2026. source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
On FourWeekMBA
- 18 Nov 2026How Anthropic’s Public Benefit Duty Works: The Trust, Class F and the $2M Suit
- 18 Nov 2026WSJ Ties Anthropic Stakes to SPVs; Filing Lists 7 Founders
- 17 Nov 2026Amazon Gets 33.9% of Anthropic’s Class N at IPO, Not 48.2%
- 15 Nov 2026Anthropic’s Benefit Fund Holds 3.5x Each Co-Founder’s Stake
- 15 Nov 2026Anthropic’s $63B Benefit Fund Is 2.3x HHMI and 68% of Gates
- 13 Nov 2026Anthropic’s 5% Public Benefit Fund Is Worth $59B-$66B at IPO
Newest 6 of 38 pages that mention Daniela Amodei.
Jared Kaplan
Chief Science Officer, co-founder, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
Career
- 2018 (paper), before AnthropicJohns Hopkins University: Associate professor of physics; taught the foundations of deep learning source
- 2018 (paper), until AnthropicOpenAI: Researcher (the 2018 batch-size paper lists him at Johns Hopkins and OpenAI) source
- Jan 2021 –Anthropic: Co-founder, Chief Science Officer
- EducationPh.D. in Physics, Harvard; bachelor's in Physics and Mathematics, Stanford source
Key papers
- A Natural Language for AdS/CFT Correlators 2011 · arXiv · theoretical physics (AdS/CFT), from his years as a physicist
- An Empirical Model of Large-Batch Training 2018 · arXiv · the gradient noise scale: how big a training batch can usefully be
- Scaling Laws for Neural Language Models 2020 · arXiv · the neural scaling laws: loss falls as a power law in model size, data and compute
- Language Models are Few-Shot Learners 2020 · arXiv · the GPT-3 paper: few-shot learning at scale
- Constitutional AI: Harmlessness from AI Feedback 2022 · arXiv · Constitutional AI: training harmlessness from AI feedback
Papers with the other co-founders before Anthropic: Dario Amodei (2018); Sam McCandlish (2018); Jack Clark (2020); Tom Brown (2020). The year is their first paper together in our set.
The S-1 calls him one of the principal authors of the neural scaling laws and says he helped pioneer Constitutional AI. source
Stake per the S-1. 242,351,170 shares beneficially owned at 30 September 2026: 239,526,270 held of record by him or by trusts over which he shares control, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 238,972,792. S-1 table footnote
Pay. Not disclosed: the S-1’s compensation tables cover the named executive officers (Dario Amodei, Daniela Amodei and the CFO, Krishna Rao). source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
On FourWeekMBA
- 15 Nov 2026Anthropic’s Benefit Fund Holds 3.5x Each Co-Founder’s Stake
- 11 Nov 2026Amazon, Google and Founders Own 41% of Anthropic at the IPO
- 10 Nov 2026Anthropic Co-Founders to Sell $250M Each, 1.4% of Stakes
- 8 Nov 2026Bores Urges Congress to Ban RSI as OpenAI Targets 2028
- 2 Nov 2026The History of Anthropic: From a $124M Research Bet to a Public S-1
- 29 Oct 2026Anthropic’s S-1 at $2T: $331B in Awards, $199B to Founders
Newest 6 of 13 pages that mention Jared Kaplan.
Jack Clark
Co-founder; leads the Anthropic Institute, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
- Public Benefit Fund initial manager
Career
- before OpenAITechnology journalist: Wrote for The Register and BusinessWeek, covering supercomputers and neural networks source
- until AnthropicOpenAI: Policy Director: engagement with governments and regulators source
- 2017 – 2024AI Index, Stanford University: Founding member source
- 2021 – 2024US National AI Advisory Committee (NAIAC): Inaugural member
- Jan 2021 –Anthropic: Co-founder; leads the Anthropic Institute; an initial manager of the Public Benefit Fund
Key papers
- The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation 2018 · arXiv · forecasting malicious uses of AI
- Release Strategies and the Social Impacts of Language Models 2019 · arXiv · the staged release of GPT-2
- Language Models are Few-Shot Learners 2020 · arXiv · the GPT-3 paper: few-shot learning at scale
- Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims 2020 · arXiv · mechanisms for verifiable claims about AI systems
Papers with the other co-founders before Anthropic: Dario Amodei (2018); Jared Kaplan (2020); Tom Brown (2020); Sam McCandlish (2020). The year is their first paper together in our set.
He writes Import AI, a newsletter on AI research. source
Stake per the S-1. 242,271,840 shares beneficially owned at 30 September 2026: 239,446,940 held of record by him or by a trust over which he has sole voting control, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 238,893,462. S-1 table footnote
Public Benefit Fund. The S-1 names four initial managers of the Anthropic Public Benefit Fund: Dario Amodei, Daniela Amodei, Jack Clark and LTBT chair Neil “Buddy” Shah. Until the board approves the Fund’s own governance, the board decides its grants. The Fund holds 846,840,960 Class N shares, 5% of Anthropic’s fully diluted shares at 30 June 2026. source
Pay. Not disclosed: the S-1’s compensation tables cover the named executive officers (Dario Amodei, Daniela Amodei and the CFO, Krishna Rao). source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
“The greatest challenge of the 21st century is to make an increasingly fast-moving technical world ‘legible’ to a large number of people.”Jack Clark, About, Import AI (read 18 Nov 2026)
On FourWeekMBA
- 15 Nov 2026Anthropic’s Benefit Fund Holds 3.5x Each Co-Founder’s Stake
- 15 Nov 2026Anthropic’s $63B Benefit Fund Is 2.3x HHMI and 68% of Gates
- 13 Nov 2026Anthropic’s 5% Public Benefit Fund Is Worth $59B-$66B at IPO
- 11 Nov 2026Amazon, Google and Founders Own 41% of Anthropic at the IPO
- 10 Nov 2026Anthropic Co-Founders to Sell $250M Each, 1.4% of Stakes
- 2 Nov 2026The History of Anthropic: From a $124M Research Bet to a Public S-1
Newest 6 of 20 pages that mention Jack Clark.
Tom Brown
Chief Compute Officer, co-founder, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
Career
Key papers
- Deep reinforcement learning from human preferences 2017 · arXiv · training agents from human preference comparisons
- Adversarial Patch 2017 · arXiv · printable patches that fool image classifiers
- Scaling Laws for Neural Language Models 2020 · arXiv · the neural scaling laws: loss falls as a power law in model size, data and compute
- Language Models are Few-Shot Learners 2020 · arXiv · the GPT-3 paper: few-shot learning at scale
Papers with the other co-founders before Anthropic: Dario Amodei (2017); Jared Kaplan (2020); Jack Clark (2020); Sam McCandlish (2020). The year is their first paper together in our set.
Stake per the S-1. 244,173,040 shares beneficially owned at 30 September 2026: 241,348,140 held of record by him or by trusts he controls, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 240,794,662. S-1 table footnote
Pay. Not disclosed: the S-1’s compensation tables cover the named executive officers (Dario Amodei, Daniela Amodei and the CFO, Krishna Rao). source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
On FourWeekMBA
- 17 Nov 20262 Days Pre-IPO, Gurley Asks Anthropic to Disclose ‘Handoff’
- 17 Nov 2026JPMorgan Cuts Nvidia’s Share of AI Lab Compute Deals to 28%
- 15 Nov 2026Anthropic’s Benefit Fund Holds 3.5x Each Co-Founder’s Stake
- 12 Nov 2026Anthropic Founders Can Sell $121B and Keep a 50.1% Vote
- 11 Nov 2026Amazon, Google and Founders Own 41% of Anthropic at the IPO
- 11 Nov 2026Anthropic’s AMD Bring-Up Took 3 Days vs 6-12 Months Before
Newest 6 of 18 pages that mention Tom Brown.
Sam McCandlish
Chief Architect, co-founder, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
Career
Key papers
- An Empirical Model of Large-Batch Training 2018 · arXiv · the gradient noise scale: how big a training batch can usefully be
- Scaling Laws for Neural Language Models 2020 · arXiv · the neural scaling laws: loss falls as a power law in model size, data and compute
- Language Models are Few-Shot Learners 2020 · arXiv · the GPT-3 paper: few-shot learning at scale
- Toy Models of Superposition 2022 · arXiv · superposition: why neurons mix many features
Papers with the other co-founders before Anthropic: Dario Amodei (2018); Jared Kaplan (2018); Jack Clark (2020); Tom Brown (2020). The year is their first paper together in our set.
His December 2018 batch-size paper, with Kaplan and Dario Amodei, lists him at OpenAI. source
Stake per the S-1. 243,383,850 shares beneficially owned at 30 September 2026: 240,558,950 held of record by trusts over which he shares control, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 240,005,472. S-1 table footnote
Pay. Not disclosed: the S-1’s compensation tables cover the named executive officers (Dario Amodei, Daniela Amodei and the CFO, Krishna Rao). source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
On FourWeekMBA
- 15 Nov 2026Anthropic’s Benefit Fund Holds 3.5x Each Co-Founder’s Stake
- 14 Nov 2026Claude Fable 5.2 Leads Epoch’s Fact Recall Test by 16 Points
- 11 Nov 2026Amazon, Google and Founders Own 41% of Anthropic at the IPO
- 10 Nov 2026Anthropic’s IPO Filing Now Says It Met Its AI R&D Threshold
- 10 Nov 2026Anthropic Co-Founders to Sell $250M Each, 1.4% of Stakes
- 6 Nov 2026Model Makers Ran 4 of This Week’s 8 AI Agent Studies
Newest 6 of 15 pages that mention Sam McCandlish.
Chris Olah
Co-founder; leads interpretability research, since January 2021 source
- Co-founder, January 2021
- Founder LLC member
- Selling stockholder
Career
- 2016 – 2017 (papers)Google Brain: Research scientist; Concrete Problems (2016) and Feature Visualization (2017) list him at Google Brain source
- 2016 – 2021Distill: Co-founder and editor of the journal source
- 2020 (paper), until AnthropicOpenAI: Led interpretability research (Zoom In, March 2020, lists him at OpenAI)
- Jan 2021 –Anthropic: Co-founder; leads interpretability research
Key papers
- Concrete Problems in AI Safety 2016 · arXiv · five research problems in AI accident risk: side effects, reward hacking, scalable supervision and more
- Feature Visualization 2017 · Distill · how a vision network builds up its understanding, layer by layer
- Zoom In: An Introduction to Circuits 2020 · Distill · the circuits programme for reading neural networks
- A Mathematical Framework for Transformer Circuits 2021 · Transformer Circuits Thread · the first Anthropic interpretability paper on transformers
- Toy Models of Superposition 2022 · arXiv · superposition: why neurons mix many features
Papers with the other co-founders before Anthropic: Dario Amodei (2016). The year is their first paper together in our set.
Stake per the S-1. 242,882,420 shares beneficially owned at 30 September 2026: 240,057,520 held of record by him, plus 2,185,400 options and 639,500 RSUs that vest within 60 days. After selling 3,378,378 in the IPO: 239,504,042. S-1 table footnote
Pay. Not disclosed: the S-1’s compensation tables cover the named executive officers (Dario Amodei, Daniela Amodei and the CFO, Krishna Rao). source
Commitment. Like each co-founder, pledged to dedicate 80% of personal equity in Anthropic to philanthropic causes; the S-1 calls the pledges voluntary and revocable, and they do not touch the Founder LLC’s votes. source
“I work on reverse engineering artificial neural networks into human understandable algorithms.”Chris Olah, About, colah's blog (read 18 Nov 2026)
On FourWeekMBA
- 15 Nov 2026Anthropic’s Benefit Fund Holds 3.5x Each Co-Founder’s Stake
- 11 Nov 2026Amazon, Google and Founders Own 41% of Anthropic at the IPO
- 10 Nov 2026Anthropic Co-Founders to Sell $250M Each, 1.4% of Stakes
- 2 Nov 2026The History of Anthropic: From a $124M Research Bet to a Public S-1
- 29 Oct 2026Anthropic’s S-1 at $2T: $331B in Awards, $199B to Founders
- 27 Oct 2026Anthropic’s Benefit Fund Holds 3.5x Any Co-Founder’s Shares
Newest 6 of 17 pages that mention Chris Olah.
The papers that tie them
20 papers, oldest first. A dot means the co-founder is an author; teal rows are before Anthropic. Links go to arXiv or the paper’s own page.
More on Anthropic
Explainers on FourWeekMBA
- How Anthropic’s Public Benefit Duty Works: The Trust, Class F and the $2M Suit
- Anthropic’s Long-Term Benefit Trust, Explained
- Anthropic Stock (ANT) and IPO: Price Range, Revenue, Losses and News, Tracked
- Who Owns Anthropic? The Founders, the Trust, Amazon and Google in the S-1
- How the Anthropic IPO Works: Price Range, Pricing Day, Listing and the Lock-Up
- Anthropic Business Model: How Anthropic Makes Money (2026)
- The History of Anthropic: From a $124M Research Bet to a Public S-1
- What is Anthropic?
- Anthropic’s Trust Network: How AI Safety Drives Structure
- The China Factor — White House Suspected Chinese Group Accessed Anthropic’s Mythos
- Anthropic’s CEO Refused the White House. 96 Hours Later, Fable 5 Was Dead.
- Anthropic in Talks for $170B Valuation, Nearly 3x in Five Months
- Anthropic vs OpenAI: The Vatican Showdown Over AI Jobs
FAQ
Who are the co-founders of Anthropic?
Seven people, per its S-1: Dario Amodei (CEO), Daniela Amodei (President and Chairperson), Tom Brown (Chief Compute Officer), Jack Clark (leads the Anthropic Institute), Jared Kaplan (Chief Science Officer), Sam McCandlish (Chief Architect) and Chris Olah (leads interpretability research). All seven worked at OpenAI before co-founding Anthropic in January 2021.
How much of Anthropic does each co-founder own?
Between 240.6 million and 244.2 million Class A shares each before the IPO, 3.6% to 3.7% of Class A; after each sells 3,378,378 shares, 3.2% each. Tom Brown has the largest holding and Dario Amodei the smallest, 3,535,060 shares apart.
What is the Founder LLC?
A company owned and managed by the seven co-founders that holds the single share of Class F stock, which carries 50.1% of the votes on stockholder matters whatever the number of shares. Each co-founder has one vote on how it is cast.
Who controls Anthropic after the IPO?
On stockholder votes, the co-founders: 61.3% including the Founder LLC. On the board, the Long-Term Benefit Trust elects 4 of the 7 directors; Class A and Class F together elect the other 3, initially Dario Amodei, Daniela Amodei and Yasmin Razavi.
Are Dario and Daniela Amodei related?
Yes. The S-1 says they are siblings, and that there are no other family relationships among Anthropic’s executive officers and directors.
Is Benjamin Mann a co-founder?
The S-1 says he is often referred to as one, but he is not a member of the Founder LLC, and the filing’s “co-founders” are the seven.
Did the co-founders sell shares?
Each sells 3,378,378 shares in the IPO. Dario and Daniela Amodei also sold in tender offers in June 2025 ($5.61 a share) and April 2026 ($25.9136); the filing does not give their share counts.
Sources and method. Stakes, votes, pay, sales, board and Trust facts come from Anthropic’s S-1/A No. 3 filed with the SEC on 17 November 2026 (ownership as of 30 September 2026). Careers come from the S-1 and from the affiliation lines of the papers themselves; authorship from arXiv records and the papers’ own pages; quotes and roles outside the filing from the person’s own site. Dollar amounts at the $70.00 to $78.00 range and all sums are our arithmetic, labelled where they appear; they are not net-worth estimates. Each figure and each line in the graph is listed with its source, exact quote and date in the page’s trace file. This page covers public, professional facts only. Not investment advice.
