认识 Anthropic 的七位联合创始人Meet the seven co-founders of Anthropic

Dario and Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, and Chris Olah each stand to make $250 million in cash proceeds and to hold stock worth upwards of $18 billion after Anthropic's IPO.

The co-founders were previously at OpenAI, ultimately leaving at the end of 2020 to start a frontier lab focused on safety.

Four researchers, a former journalist, a public policymaker, and an engineer founded Anthropic in 2021. Six years later, the artificial intelligence company is about to go public, with trading expected to begin on Nov. 19, in what will be one of the biggest IPOs on record. Those seven co-founders are poised to each be worth many billions of dollars. It all started with a defection.

At the end of 2020, siblings Dario and Daniela Amodei were vice presidents at OpenAI the company that kicked off the generative AI frenzy with the launch of ChatGPT. But the two had grown frustrated with their employer over its approach to AI safety. Fearing a world where models' capabilities would progress faster than our understanding of them, the duo launched Anthropic early the next year.

They took a handful of OpenAI colleagues with them: Tom Brown, the company's research engineering lead on GPT-3; Jack Clark, a policy director; Jared Kaplan, a contract research consultant; and Sam McCandlish and Chris Olah, both members of OpenAI's technical staff.

They were all deeply concerned about OpenAI's ambitions to focus on commercialization, and they now stand to be some of the first to cash in on the AI boom.

"They really started Anthropic not because they were hoping to get rich, but they started Anthropic because they like wanted to shape this really important thing in the world, and they thought that other people might do it worse than them," said Jeffrey Ladish, who joined Anthropic as a contract safety engineer in September 2021, and now serves as the executive director of AI safety institute Palisade Research.

Collectively the co-founders own about 10% of Anthropic. Each has about 240,000,000 class A shares, or about a 1.4% stake.

During Anthropic's most recent funding round, those shares — valued at $58.90 each — amounted to about $14 billion in value for each co-founder. If Anthropic's IPO reaches its target of $70 to $78 per or a market valuation of nearly $1.4 trillion, it would push those stakes to $17 billion to $18 billion.

"It's extremely hard not to be influenced by your incentives," Ladish said. "You have investors, and money is not just like an end; it's also a means." Anthropic did not respond to requests for comment.

Dario Amodei The face of Anthropic is CEO Dario Amodei. He holds the most shares of the bunch: 245,259,100 including shares subject to voting proxies.

Amodei, 43, didn't get his start in business. He comes from the world of research, obtaining a doctorate in computational neuroscience from Princeton University in 2011. He then went to work at Stanford University as a postdoctoral scholar from 2011 to 2014.

Even in the private sector, Amodei's experience was focused on research. He worked at China's Baidu leading research on speech recognition from 2014 to 2015, followed by a stint at Google Brain working on deep learning and neural networks, foundational scientific frameworks for today's large language models. He joined OpenAI in July 2016.

On the Big Technology podcast in July 2025, Amodei said he got into AI to solve the "almost intractable" and "slow moving" problems of biology.

"I think my reason to start a company was that I had worked at other companies and I just didn't feel like the way those companies were run was really oriented towards trying to have that impact," he said, adding that the desire for impact was often used by companies as an insincere ploy to recruit talent.

At OpenAI, Amodei is credited with co-inventing reinforcement learning from human feedback (RLHF), alongside Anthropic co-founder, Tom Brown. RLHF is a fine-tuning method used to align model outputs with human needs. That could mean training a model to speak politely or to not offer violent or harmful solutions to problems.

Daniela Amodei Four years her brother's junior, Daniela Amodei, 39, serves as the president and board chair of Anthropic. The Amodeis are both San Francisco natives, growing up in the Mission District, just a few miles away from Anthropic's current headquarters.

The pair were teenagers during the height of the dot-com bubble and its subsequent fallout, both attending Lowell High School, a magnet school and one of the most prestigious public schools in the country. After graduating, Daniela went on to study English literature and politics at the University of California, Santa Cruz, earning her degree in 2009.

Daniela started her career in politics, working for Matt Cartwright, a Democrat representing what was then Pennsylvania's 17th congressional district, and then moved into tech, taking on various roles in risk management at Stripe for five years.

In a June interview with Stanford's Graduate School of Business, she said her experience in politics, specifically campaigns, naturally led her to startups and Silicon Valley.

"I was like, 'Wow, a small number of people that work really hard, who are young and driven, can really change the world,'" she said. Tom Brown Anthropic's chief compute officer, Tom Brown, 39, is perhaps the most entrepreneurial of the bunch. After graduating from the Massachusetts Institute of Technology with a Master of Engineering in 2010, Brown went on to found two startups, MoPub and Grouper.

MoPub was a mobile advertising business acquired by Twitter in 2013 There he served as founding engineer in 2011. Grouper, where Brown served as co-founder and technology chief, was a Y Combinator-backed group dating app.

After Grouper, Brown pivoted to research, sharing on Y Combinator's Lightcone podcast in August 2025 that he was self-taught on machine learning research, a daunting task since he received a B- in linear algebra at university. "I kind of flip-flopped on it for six months trying to build up courage to" make the switch, he said.

Alongside co-inventing RLHF, as a researcher at OpenAI, Brown led the training efforts on GPT-3, a model that marked a historic leap from GPT-2, going from 1.5 billion parameters to 175 billion.

Jack Clark Jack Clark, 37, the head of public benefit at Anthropic, left his career as a journalist in 2016 to work at OpenAI. Clark is now also the leader of the Anthropic Institute, the company's internal think tank that explores AI's social and economic impacts.

A native of the U.K., Clark grew up in the seaside town of Brighton and later studied English literature and creative writing at the University of East Anglia. He then went on to write for publications including ZDNet UK, The, and Bloomberg.

In an email to a colleague at Bloomberg at the time, Clark said his decision to leave journalism was because he had "never believed in something strongly before" and that he needed to join OpenAI "and live with the consequences, which I hope are good." Jared Kaplan Kaplan, 42, Anthropic's chief science officer, still holds a position as an associate professor in the department of physics and astronomy at Johns Hopkins University. The theoretical physicist got his PhD in physics from Harvard in 2009, working for four years after as a postdoctoral fellow at the SLAC National Accelerator Laboratory, a particle physics research facility run by Stanford and funded by the Department of Energy.

Kaplan co-pioneered scaling laws, or the notion that large language models get exponentially better when trained on larger and larger data sets, alongside Dario Amodei. He also pioneered "constitutional AI," or a method of aligning models by using a list of safety principles and policies. His day-to-day work concerns safety assessments for models, responsible scaling policy at the company, and existential risks.

interview with The Guardian in December 2025, Kaplan noted that the decision over whether or not to let AI "recursively self improve," or write the code to enhance itself without a human in the loop, would be "in some ways the ultimate risk, because it's kind of like letting AI kind of go," and that he expects this risk will be taken sometime between 2027 and 2030.

Sam McCandlish Anthropic's chief architect, Sam McCandlish, 36, came to OpenAI as a fellow focused on AI safety. He also led early research on coding tool Codex, which now constitutes a significant portion of OpenAI's revenue.

McCandlish claims he and the rest of the Anthropic co-founders were not following an entrepreneurial spirit when they charted out on their own. In a December 2024 video from Anthropic's YouTube channel, he said, "None of us wanted to found a company. We just felt like it was our duty." McCandlish, like Kaplan, also came from the world of academia, completing his PhD in theoretical and mathematical physics at Stanford in 2017 and working at Boston University as a postdoctoral fellow the year afterward. His work focused on quantum gravity and tensor networks.

Now, his work focuses on pre-training, or the first step of training a large language model on swaths of data, among other things.

Chris Olah Chris Olah, 34, heads Anthropic's interpretability research, a branch of study in machine learning that attempts to understand how a large language model "thinks" — or rather, which data inputs and reasoning methods it uses to arrive at its outputs.

Like McCandlish, Olah was not originally keen on starting a company. In the same December 2024 YouTube video, Olah said, "I was arguing for a long time that we should do a nonprofit instead, and just focus on safety research." Olah had a nontraditional path to Anthropic, dropping out of the University of Toronto after one year and working for short stints as a developer at the government agency Environment Canada and software startup Xelerance.

His career truly took off when he was awarded a Thiel Fellowship in 2012, which he used to pursue research pertaining to 3D printers for two years. He then joined Google Brain, first as an intern and then as a researcher, before moving to OpenAI in 2018.

Majority voting control While together the seven founders may only own about 10% of total equity, they will have a majority vote.

Founder LLC the legal entity set up that owns a single Class F share, gives Anthropic's co-founders a 50.1% majority voting control over the company. While Benjamin Mann, a member of Anthropic's technical staff and also an ex-OpenAI employee, is often called a co-founder, he is not an owner of Founder LLC.

The co-founders are expected to collectively sell over 20 million shares during the IPO, resulting in roughly $250 million in cash per founder, according to Anthropic's prospectus.

They plan to dedicate 80% of the wealth from their personal Anthropic equity to philanthropy, according to the IPO filing, though the pledge isn't legally binding. Anthropic is also setting aside 5% of its shares to form a new public benefit fund— to finance nonprofits, research, and potentially direct community grants — with potential to grow the size of the fund as the company itself grows. Full interview with Anthropic co-founder Daniela Amodei