Later this month, traders are likely to get a chance at the first AI mega-IPO, as Anthropic sells shares to the public. Investors will get a chance to participate in what Anthropic’s public filing calls a $40 trillion opportunity, profiting from possibly the biggest technological advance of all time.
Those investors won’t, however, get a say in how the company is run. Since the beginning, Anthropic’s leadership has presented its mission as too important to be subject strictly to the profit motive. Its SEC filing lays out an elaborate corporate structure—one ostensibly designed to keep Anthropic focused on the public interest but that in effect leaves CEO Dario Amodei and his co-founders firmly in control.
Finally, a reason to check your email.
Sign up for our free newsletter today.
Over the last two decades, founders of tech companies increasingly have sought to maintain control of their companies as they reap the financial benefits of going public. The original, fairly modest goal was to ameliorate Wall Street’s focus on short-term results. Google pioneered a structure that gave its founders super-voting shares in 2004. Since then, other founders, from Mark Zuckerberg at Meta to Elon Musk at SpaceX, have adopted the approach, often adding levers to make sure they stay in charge.
Anthropic has taken this strategy a step further. Rather than simply giving its founders super-voting shares, its proposed corporate structure employs nesting-doll governance, with its board of directors selected by a series of interlocking committees. Anthropic’s founders retain 50.1 percent of shareholder voting rights through their control of a special Class F share. That gives them the ability to name three of the company’s seven directors. Three more directors are chosen by an ostensibly independent Long Term Benefit Trust, a self-perpetuating oversight board that votes on its own members. The seventh and final director is also chosen by the Trust, but with effective veto power held by the founders.
Then there are the Class E shares, which are given to employees mainly in proportion to the shares of stock that they pledge to donate to charity. The founders plus a supermajority of the Class E shares, together with (finally—the ordinary owners!) a majority of regular investors’ Class A shares can vote to disband the Trust.
The reason for this structure may be the fact that Amodei, like each of Anthropic’s other founders, owns only about 2 percent of the company’s shares. The Long Term Benefit Trust gives him and his fellow founders control wildly disproportionate to that ownership stake. New York University Stern School of Business Professor Aswath Damodaran, one of the world’s leading experts on how to value companies, calls this structure “so deviously self-serving that it would put Kim Jong Un to shame.”
The justification here appears to be that mere shareholder self-interest isn’t sufficient to control a technology as protean as AI.
Amodei develops this idea in “The Adolescence of Technology,” an essay about AI progress and its risks. Amodei writes about the importance of “ensuring that AI development remains accountable to the public interest, not captured by any particular political or commercial alliance.” Similarly, Anthropic’s charter calls for the company to “develop and maintain advanced AI for the long-term benefit of humanity.”
In theory, putting control in the hands of a committee responsible to the public good binds Anthropic to that mission. In practice, there’s nothing to suggest Anthropic’s governance structure will actually keep the company oriented toward the “long-term benefit of humanity.”
We know this from the example of a similarly structured company where everything went wrong: Anthropic’s chief competitor, OpenAI. OpenAI used to be controlled by an independent nonprofit board. Three years ago, the independent nonprofit board that oversaw OpenAI sought to depose CEO Sam Altman in a public coup. Today, Altman remains CEO, the company has been restructured, and three of the four directors who tried to oust Altman are gone.
In choosing the members of Anthropic’s Long Term Benefit Trust, Amodei is unlikely to permit a similar outcome. Rather, the Trust is a classic blue-ribbon commission. Its four current members are a mix of scientists and policymakers, including former Federal Reserve Chairman Ben Bernanke. None has an extensive track record with AI.
To be clear: artificial intelligence needs meaningful oversight. That’s especially true given its real-world risks in areas like biology or cybersecurity. The concerns voiced by Amodei and others at Anthropic about the potential for AI exceeding our abilities to control it are worth serious consideration.
But Anthropic’s corporate structure doesn’t add meaningful oversight, just the illusion of process. As Damodaran puts it, “if you have had bad experiences with top management at companies using corporate responsibility and ESG as cover for self-serving activity and hypocrisy, the Anthropic structure is that structure on steroids.”
Anti-capitalists love to caricature CEOs as cartoon villains who consider nothing and no one beyond how many dollars their corporation will earn tomorrow. The truth is that large corporations have always been answerable to public pressure, whether from institutional investors, government, or even the press. And the principle of shareholder control has the virtue of simplicity.
The structure now favored by tech companies forgoes that simplicity but compensates by adding personal accountability. You may dislike what Musk does at SpaceX or Zuckerberg does at Meta, but you will never feel lost about whom to blame.
Anthropic’s structure lacks both the simplicity of shareholder control and the personal accountability that founders with super-voting shares can’t avoid. In their place, it provides a way for founders like Amodei to take whatever path they want while claiming that independent appraisers of the public interest have vetted their actions.