MARKET INSIGHT · EARNINGS INTELLIGENCEPolaris Earnings CalendarTrack market technicals, follow the S&P 500 outlook, stay ahead of company earnings, and explore crypto technicals—all in one place.

POLARIS EARNINGS CALENDAR

Anthropic’s safety trust faces its first major test as founders retain decisive voting power

Anthropic’s safety trust faces its first major test as founders retain decisive voting power

Anthropic is preparing to enter public markets with a governance structure designed to protect its mission of developing advanced artificial intelligence for humanity’s long-term benefit, while leaving its seven co-founders with decisive control over ordinary shareholder votes. The arrangement places a financially disinterested Long-Term Benefit Trust at the center of oversight, but limits its role largely to board composition rather than daily management or strategy.

Anthropic’s structure combines Delaware public benefit corporation status with several classes of stock. The company’s Class F share, held through a founder-controlled limited liability company, carries 50.1% of the total voting power on matters submitted to shareholders generally, including the election of directors outside the trust’s designated seats. The founders are Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish and Christopher Olah, who initially hold the Class F share collectively.

The Long-Term Benefit Trust holds a single Class T share with no economic rights. Anthropic says the trust is intended to have five financially disinterested trustees with expertise spanning AI safety, national security, public policy and social enterprise. Its Class T stock generally allows it to elect and remove directors who together represent a majority of the authorized board, although the trust and founder-controlled entity collectively have the ability to elect all directors under the company’s stockholders’ agreement.

The trust does not directly run Anthropic. The company’s governing documents say the board makes decisions about the business and affairs, while the trust selects directors and provides advice and guidance. Anthropic is classified as a controlled company under Nasdaq rules and can opt out of some governance requirements; it says it expects to rely on an exemption concerning its nominating and corporate governance committee, which will not be entirely independent.

The practical limits of the model have become more visible as Anthropic’s approach to AI safety has faced criticism from Microsoft AI chief Mustafa Suleyman. Suleyman has argued that models should not be trained to present themselves as conscious beings with rights or welfare interests, while Anthropic’s published material treats questions about Claude’s identity, consciousness and possible welfare as matters requiring caution. The disagreement does not establish that either approach is correct, but it illustrates the kind of substantive safety dispute that could test whether mission-based oversight produces a public challenge or remains an internal process.

Anthropic itself described the trust as an experiment when it introduced the arrangement in 2023, saying its effectiveness would have to be observed over time. Those commitments may reinforce the company’s stated purpose, but they do not give ordinary shareholders a conventional mechanism to replace the founders or determine how broad claims about public benefit should be interpreted.

That distinction matters because the company’s stated mission is difficult to measure in the same way as revenue or profit. A public benefit corporation must balance shareholder interests, the interests of people materially affected by its conduct and its specified public purpose, but Anthropic’s documents leave substantial room for the board to judge how those interests should be balanced. As the company moves toward public ownership, greater financial disclosure may give investors more information, yet the core allocation of power remains unusual: founders control ordinary voting matters, while the trust’s principal leverage is the appointment and removal of a majority of directors. Whether that arrangement supplies meaningful accountability will depend on how the trustees act when their interpretation of AI safety diverges from the founders’ judgment.

Read the full account of the story

More from the AI desk

More from AI · Back to Polaris Earnings Calendar