Ordinary shareholders will have unusually little say in the governance of Anthropic when it goes public. The firm argues that artificial intelligence is too consequential a technology to be governed by ordinary commercial incentives. To ensure that the public interest is protected, the business says it believes its seven founders are “distinctly equipped to be stewards of our mission.”
The arrangement laid out in a Monday filing with the Securities and Exchange Commission is complicated, but the upshot is straightforward: The co-founders will control a special share carrying 50.1 percent of the voting power on most matters. A separate Long-Term Benefit Trust, whose five financially disinterested trustees largely appoint their own successors, will generally select a majority of the company’s board. Between them, they will essentially control the selection of every director. The S-1 filing says Anthropic will be considered a “controlled company” under Nasdaq rules.
There’s nothing inherently wrong — or novel — about trying to insulate management from shareholders. Investors are free to buy shares on those terms — or not. Silicon Valley has a long tradition of asking Wall Street to place its faith in visionary founders. Google’s 2004 prospectus famously enshrined a “Don’t be evil” principle and warned that anyone buying the stock was making a long-term bet on Larry Page and Sergey Brin. Mark Zuckerberg retains extraordinary control over Meta.
But Anthropic makes a grander claim for its governance structure. It says this concentration of authority is good for humanity. That deserves considerably more skepticism.
Anthropic’s view is that AI is unique from anything that has come before. Its prospectus describes a technology that could deliver extraordinary scientific advances and material benefits to humanity but could also cause widespread displacement of workers and even escape human control.
Its answer to these concerns does not instill confidence: Anthropic argues that markets aren’t particularly good at accounting for such outsize externalities. The firm’s solution is to entrust a small group of founders and trustees with extraordinary discretion.
The great insight of the economist Friedrich Hayek was that knowledge is dispersed throughout society, not bequeathed to the enlightened few. Markets may be imperfect, but competition, experimentation and the possibility of failure provide sounder mechanisms for correcting mistakes than the judgment of any individual.
Something funny happens whenever enlightened technocrats try to make capitalism more “benevolent”: From corporate social responsibility to stakeholder governance and now public-benefit corporations, the solution usually involves giving those in charge greater discretion to decide what constitutes the public good.
Anthropic needs to remember it is a corporation, not a government. It’s reasonable for the public to wonder why a business needs to insulate itself to act in the public’s purported interest. As AI develops, and if new risks emerge, democratically elected governments may assert themselves. For now, market discipline and enforceable liability offer imperfect but valuable checks on the behavior of the frontier labs.
Who guards the guardians? The Roman poet Juvenal asked that question nearly 2,000 years ago. The appropriate response to a future that nobody can confidently predict is not to concentrate more power in the hands of a clique of self-appointed guardians.
The post Anthropic’s plan to protect humanity has a flaw appeared first on Washington Post.
