Anthropic, an artificial intelligence company preparing for its initial public offering (IPO), has revealed plans to allocate 5% of its shares to a public benefit fund aimed at supporting nonprofit organizations and advancing its broader mission. The company stated it intends to increase this allocation if the fund’s value surpasses a predefined threshold. According to Anthropic, the fund will not be involved in political donations or lobbying activities.

However, an influential yet little-known element of Anthropic’s governance remains largely undisclosed: the Long-Term Benefit Trust, which holds significant control over the company’s board and certain major corporate decisions. Unlike most public companies where investors have primary influence, Anthropic is structured as a public-benefit corporation (PBC) governed in part by this trust, which is vested with substantial oversight responsibilities. The trust appoints four of the seven members of Anthropic’s board, has the authority to remove board members without cause, and can block critical corporate actions. Its stated mission is to guide the development of AI technology for humanity’s long-term benefit.

The trust’s members include notable figures such as former Federal Reserve Chairman Ben Bernanke and Neil Buddy Shah, chairman of the Clinton Health Access Initiative. According to Anthropic, these trustees are financially disinterested, hold no equity in the company, and do not share in its profits, being compensated solely for their time and service. The trust is required to consult with company leadership on certain key issues, though the precise nature of these consultations has not been disclosed.

Anthropic’s legal advisers have detailed that the company must notify the trust at least five days in advance before undertaking significant actions such as issuing substantial new debt, selling a controlling interest, or transferring control over AI models that meet specific performance milestones. Some legal experts interpret this notification requirement as an implicit veto, given the trust’s ability to reconstitute the board majority. Anthropic denies the trustees hold veto power but declined to clarify the mechanics further.

The rise of public-benefit corporations over the past decade and a half reflects a growing trend in corporate governance, allowing firms to balance shareholder interests with broader societal or environmental considerations. Delaware law, under which Anthropic is incorporated, mandates PBC directors to weigh the interests of those materially affected by the company’s operations alongside shareholders’ interests and the corporation’s stated public benefit goals.

Anthropic is among a small group of PBCs preparing to go public, joining other firms like Warby Parker and Veeva Systems, as well as competitor OpenAI, which is controlled by a Delaware-based nonprofit. Scholars caution, however, that the mission of a PBC may be vulnerable over time if directors prioritize financial returns, given they are elected by shareholders who stand to benefit from profit maximization.

Anthropic asserts that the Long-Term Benefit Trust exists specifically to maintain fidelity to its mission and to carefully balance shareholders’ financial interests with broader societal concerns influenced by its operations. Nonetheless, corporate governance experts note that this structure limits the influence of regular shareholders, who will hold few rights relative to the trust’s authority once Anthropic is publicly traded.

According to a 2023 statement, changes to the trust’s powers require approval by “sufficiently large supermajorities” of shareholders, though the exact thresholds and conditions remain unclear. The company’s latest incorporation documents suggest that eliminating or diluting the trust’s role would require its consent, along with support from a substantial portion of the board or a shareholder majority potentially as high as 85%.

Despite these governance complexities, Anthropic is testing market appetite for its IPO, which could raise up to $100 billion and value the company near $2 trillion, signaling strong investor interest in the AI sector.