Anthropic, the artificial intelligence company behind the Claude language model, is preparing for a potential initial public offering (IPO) that could value the firm at up to $2 trillion, bringing heightened attention to its unique governance structure. Central to this setup is the Long-Term Benefit Trust (LTBT), a small group of trustees that wields significant control over the company’s board despite holding no equity. The trust was established to ensure that Anthropic’s mission—to develop AI for the long-term benefit of humanity—is safeguarded even as commercial pressures increase.

Headquartered in San Francisco, Anthropic operates as a public benefit corporation, with the LTBT possessing the right to appoint or remove the majority of its board members. Currently, the trust has appointed four of the seven directors, including Netflix cofounder Reed Hastings and Novartis CEO Vas Narasimhan. The LTBT’s three members include Neil Buddy Shah, chief executive of the Clinton Health Access Initiative and chair of the trust, former Federal Reserve chair Ben Bernanke, and Richard Fontaine, CEO of the Center for a New American Security. Mariano-Florentino “Tino” Cuéllar, a former California Supreme Court justice, briefly served on the trust before transitioning to Anthropic’s chief global affairs officer role.

Anthropic’s governance model aims to set a precedent for the AI industry, providing a blueprint that balances innovation with ethical oversight. The trustees are kept well-informed of major company developments, including new AI model launches and strategic decisions. They meet internally on a weekly basis and engage frequently with Anthropic’s leadership, discussing significant matters such as the controlled deployment of the Mythos cybersecurity model and the company’s dealings with the U.S. government concerning automated weapons.

Despite the trust’s extensive involvement, insiders say it has mostly played an advisory role and has yet to confront a major conflict between profitability and the company’s societal goals. This arrangement has drawn criticism from governance experts who warn of inherent tensions. Jesse Fried, a Harvard Law professor, points to a “built-in conflict” in Anthropic’s structure: while the company attracts profit-focused investors, the trustees have a mandate to prioritize its mission, potentially leading to difficult trade-offs. Elizabeth Pollman of the University of Pennsylvania highlights that unpredictable challenges are difficult to contract for, especially amid intense geopolitical and commercial competition in the AI sector.

Anthropic’s approach contrasts with OpenAI, another leading AI lab that also employs nontraditional governance with self-appointed mission guardians rather than traditional directors. OpenAI’s governance came under scrutiny last year following a board attempt to remove its CEO, Sam Altman, which ultimately failed and triggered broader organizational upheaval. Unlike OpenAI, Anthropic’s governance includes a "kill switch" provision that allows shareholders holding 85 percent of voting power to remove the trustees, a mechanism seen by some as a limiting safeguard.

Private investors backing Anthropic have reportedly endorsed its governance model and have cited the company’s safety focus as a key investment criterion. However, some stakeholders acknowledge an expectation that Anthropic would still need to become a commercial powerhouse to support its ambitious mission. As the company moves toward public markets, it will face new pressures from a wider investor base demanding profitability.

Experts advise close scrutiny from potential investors, emphasizing that Anthropic’s governance structure remains largely untested under the strain of balancing commercial and ethical imperatives. While seen as less vulnerable than OpenAI’s model, questions remain whether the trust can effectively navigate competing interests as the company scales. The coming months will be critical in determining whether Anthropic’s governance framework can serve as a viable template for the rapidly evolving AI industry.