The Anthropic Long-Term Benefit Trust

Key components of the Anthropic Trust

Anthropic, PBC, one of America’s leading AI labs, is pioneering an innovative approach to corporate governance for the development of artificial intelligence. They established the “Anthropic Long-Term Benefit Trust,” a novel arrangement that empowers experiments in artificial intelligence and ethics to gradually pick a majority of Anthropic’s board of directors. In a paper from the Harvard Law School Forum on Corporate Governance, John Morley, a professor at Yale Law school, David J. Berger and Amy L. Simmerman, partners at Wilson Sonsini Goodrich & Rosati, make an outline of some of the key components of the Trust, as well as the motives behind adopting it. Here is a summary!

Photo by Zhanhui Li on Unsplash

The Origins: The Anthropic Long-Term Benefit Trust, according to Anthropic, emerged from a deep commitment to promoting social good. Anthropic’s founders recognize and acknowledge the great potential behind AI and also believe that the safety and social benefits of it are closely tied to profitability and commercial success. Anthropic also aimed to design a sort of legal framework that would ensure safety and responsibility all while facilitating investor profits. 

As mentioned in the Harvard Law paper, the company’s first step was organizing itself as a Delaware public benefit corporation (PBC), which would allow its board of directors to balance the interests of stockholders with the aim of its mission: “responsibly develop and maintain advanced AI for the long-term benefit of humanity.” Their innovative approach involves supplementing the PBC structure with the Anthropic Long-Term Benefit Trust, built around a special class of shares called Class T Common Stock. The trust operates under its own governance structure as a common law trust governed by Delaware law. Different from your typical trusts, it is a purpose trust that is managed for achieving a specific mission rather than benefiting particular beneficiaries. This mission aligns with the company's goal of combining profit with responsible AI development. 

Internal Governance: The internal governance of the trust was designed to strike a balance between the independence of Trustees and a working relationship with the company. The initial trustees were chosen by the company; however, in the future, they will be chosen by existing Trustees. Furthermore, Trustees maintain working connections with the company because of provisions in the Trust’s documents that require consultation. Although the trustees can request any appropriate information or resources, the company can withhold information or resources for certain purposes as well.

Finally, the trustees’ independence is further balanced by accountability. This means Trustees must consult with the company directors and CEO regarding appointments of Trustees. To ensure re-evaluation, they also only serve one-year terms.

The process for amending the Trust can begin through the consent of the Voting Trustees and the company’s stockholders or with the consent of the Voting Trustees or otherwise. Amendments can also be made by a supermajority of stockholders. Their last option that does not require permission of the Voting Trustees, instead acts as a fail-safe mechanism against potential actions by the Trustees. 

Read more about it here!