Anthropic’s seven founders tighten grip ahead of blockbuster IPO

{"text":[[{"start":10.32,"text":"Anthropic’s seven billionaire co-founders are set to gain collective control over one of the world’s biggest companies following its blockbuster listing, testing whether shared beliefs and personal bonds can withstand the pressures of the public market."}],[{"start":25.08,"text":"The AI lab issued its IPO prospectus on Monday, revealing that seven executives — chief executive Dario Amodei and president Daniela Amodei, alongside Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah — will together gain a special class of stock that grants them 50.1 per cent of the voting power."}],[{"start":47.6,"text":"The move means ordinary shareholders will have virtually no control over Anthropic’s decisions. Instead, investors must bet on a group that is described as “stewards of our mission” in the S-1 prospectus. The document pointed to the “uncommon” retention of all seven — a subtle critique of OpenAI, Anthropic’s rival where the majority of co-founders have left."}],[{"start":null,"text":"

"}],[{"start":69.08,"text":"“The Anthropic founding team has an incredibly unified vision and culture, which is something that is pretty rare among frontier AI labs,” said Jenny Xiao, former OpenAI researcher and partner at Leonis Capital."}],[{"start":81.84,"text":"“This is pretty reassuring for investors because even though the company is pretty established, leadership turnover still takes away institutional knowledge and hurts cultural coherence.”"}],[{"start":92.96,"text":"While the company’s valuation by public investors will determine the net worth of the co-founders, their stakes were estimated to be worth up to $15.5bn each after Anthropic’s last fundraising valued it at $965bn in May. All seven have pledged to donate 80 per cent of their wealth to charity."}],[{"start":111.72,"text":"Anthropic’s seven co-founders were colleagues at OpenAI until December 2020, when they quit to start their own lab. A shared belief spurred them on: AI systems were scaling rapidly, and rolling them out required a more responsible path."}],[{"start":126.16,"text":"“Things are never going to be chill again,” Amodei told an early company meeting."}],[{"start":130.84,"text":"He gathered the seven to split off to form their own organisation. The idea was first suggested by Kaplan, a friend from graduate school, according to people familiar with the matter. Amodei had previously worked with Brown and Olah at Google."}],[{"start":null,"text":""}],[{"start":144.88,"text":"Kaplan’s early work at OpenAI included research with McCandlish and Amodei on “scaling laws”, or how AI model capabilities increase with data and computing power. This research has shaped the AI sector’s current approach to developing models."}],[{"start":160.52,"text":"Daniela Amodei, who worked on AI safety and policy at OpenAI, and Clark, a former Bloomberg journalist who worked on OpenAI’s policy and communications, also moved to start Anthropic."}],[{"start":172.24,"text":"“The seven of us really felt like we had our own vision, and we wanted to just go do that our own way, somewhere else, because it was less friction,” Daniela Amodei told the FT earlier this year."}],[{"start":184,"text":"The co-founders questioned whether to form a company at all, or whether to form an organisation styled on Cern, the renowned Swiss physics centre, according to people familiar with the matter. Many within the group expressed doubts about creating a for-profit venture, they added."}],[{"start":null,"text":""}],[{"start":199.76,"text":"Ultimately, they recognised the need to raise a huge amount of capital to run AI models and settled on creating a public benefit corporation, a company required to consider its impact on society alongside shareholder returns."}],[{"start":213.28,"text":"The co-founders remain extremely influential at Anthropic and personally tight, according to people close to them, who said they continue to regularly meet and share takeaways from the same Chinese restaurant in San Francisco they have frequented for years. They share other points of commonality: several have had children since founding the company and many are — or have been — vegetarian."}],[{"start":235.16,"text":"The Amodei siblings are at the helm of Anthropic, splitting core responsibilities. Dario sets the vision, such as by penning long public essays to guide the company, while diving deep into research. He has only one direct report, his chief of staff."}],[{"start":250.48,"text":"Daniela, however, runs the organisation day to day. She oversees the rest of the executive team and operations, from research and engineering to policy and communications. Internally, she is considered a good balance to her brother, with better “people skills” and emotional intelligence, according to multiple people familiar with the siblings."}],[{"start":null,"text":""}],[{"start":269.22,"text":"The seven co-founders have roughly an equal number of shares, according to the S-1 filing, and all retain key executive functions. Clark has taken on a public-facing policy role, as the self-proclaimed “AI Wikipedia”, using his skills as a former journalist to translate complex technical work and thorny AI ethical issues to government officials and the public."}],[{"start":291.56,"text":"Olah, the interpretability research lead who focuses on the technology and the scientific work underpinning it, recently spoke at the Vatican during the presentation of Pope Leo XIV’s AI encyclical, an influential letter in which the pontiff called for AI to be “disarmed” through regulation and limits on the warfare."}],[{"start":null,"text":""}],[{"start":309.8,"text":"Brown, Anthropic’s chief compute officer, helps broker access to computing capacity and is known to have rescued relations between Anthropic and the Trump administration over restrictions on the company’s models. That intervention came after conversations with the government were hampered by a personality clash between Dario Amodei and Pentagon official Emil Michael."}],[{"start":331.28,"text":"Benjamin Mann, who leads experimental work at Anthropic, was the eighth person to join the company and is often described as a co-founder, but is not part of the voting block, according to the S-1."}],[{"start":342.36,"text":"The seven co-founders will set the company’s future. At the top of Anthropic’s unusual governance structure is a “Long-Term Benefit Trust”, which appoints its own members and generally has the power to elect and remove four out of the seven directors on its board."}],[{"start":356.4,"text":"The trust’s control over the fourth board seat — the casting vote — is limited by a requirement that the board’s nominating committee, which includes Daniela Amodei, agrees with its decisions. Meanwhile, neither Dario nor Daniela can be removed from Anthropic’s board by the LTBT."}],[{"start":372.48,"text":"The S-1 states the co-founders will form a Founder LLC which will, in effect, have the votes to appoint three members of Anthropic’s board."}],[{"start":380.88,"text":"Their controlling interest will also allow them to have a decisive say on all significant decisions, such as future acquisitions. Within the Founder LLC, each co-founder will have an equal vote, with a majority needed for most decisions to be approved."}],[{"start":395.12,"text":"Those requirements require continued unity between the co-founders for the company to function well."}],[{"start":400.16,"text":"“They may share a commitment to safe AI but still disagree about when to release a model, how much to spend or which markets to enter,” said Song Ma, professor of finance and entrepreneurship at Yale School of Management."}],[{"start":413,"text":"“Those differences would have consequences for both management and control. How these people handle their internal disagreement will be super important as opinions on AI in the general public start to diverge.”"}],[{"start":424.76,"text":"Additional reporting by Rachel Rees in London"}],[{"start":432.56,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1793176604_6147.mp3"}