Anthropic’s IPO limits shareholder control over the AI lab

By Sophie Lindqvist

Anthropic’s planned public-market structure gives its seven founders 50.1% of shareholder voting power and lets a safety trust elect a majority of the board.

For Anthropic, the coming initial public offering separates financial ownership from corporate control. Anthropic founded the lab in 2021 to ensure humanity can control AI and benefit from it in the long term. Public shareholders will own stock without gaining much influence over major decisions, including who joins the board.

This matters because Anthropic is putting AI safety decisions under a structure that shields them from ordinary shareholder voting. Public investors gain a financial stake in that model, but not the final word over the people accountable for it.

What kind of company is Anthropic becoming?

Anthropic is a public benefit company: a for-profit corporation that can weigh a public mission alongside profit. Its mission is AI safety. That status protects the board’s ability to make safety decisions that aren’t driven entirely by shareholder value, without leaving it exposed to investor lawsuits.

The design resembles familiar Silicon Valley arrangements in which public-company founders retain control without owning most shares. Elon Musk and Mark Zuckerberg are the examples cited for that broader pattern. Anthropic adds layers around its founders, including a mission-based corporate form, special voting stock, an independent trust and an employee vote.

In plain terms, Anthropic is asking investors to fund a safety-focused company without giving shareholders ordinary control over its safety policy.

Who controls Anthropic’s board after the IPO?

Two special voting blocks sit behind Anthropic’s board. The first belongs to the company’s seven founders. Together, they own a special class of stock that gives them 50.1% of shareholder voting power and lets them elect three board members.

The founders’ voting power depends on unity. If no majority exists among them, their shares get voted in proportion with common stock. A split among the seven would therefore give ordinary shareholders more influence than the headline voting percentage suggests.

The second block belongs to Anthropic’s Long Term Benefit Trust, or LTBT. This independent panel exists to keep the company focused on its mission to benefit humanity. It owns stock that can elect and remove a majority of Anthropic’s board, a group known as Class T directors.

Anthropic’s current board includes:

• Daniela Amodei, the company president and board chair
• Dario Amodei, CEO and Daniela Amodei’s brother
• Reed Hastings, Netflix cofounder
• Christopher Liddell, former Microsoft and GM CFO
• Vas Narasimhan, Novartis CEO
• Yasmin Razavi, Spark Capital partner

The board has one vacancy.

The LTBT has a separate roster:

• Neil Shah, its chair, a global health expert and former managing director at GiveWell, a nonprofit that ranks charities
• Ben Bernanke, former Federal Reserve chair
• Richard Fontaine, national security expert
• Christopher Stubbs, Harvard physicist

The trust has one vacancy, and its existing members choose who fills it.

The LTBT and Anthropic’s board remain separate bodies. The trust supplies the voting power to select a majority of directors, while the corporate board oversees the company. Investors therefore face two layers of special control before their common votes have any effect.

Shareholders will own Anthropic, but the founders and the LTBT will control who governs it.

What power can employees take?

The corporate board can check the LTBT in a narrow but consequential way. When the LTBT nominates a “Specified” one of its four directors, or removes that director without cause, the board’s nominating committee can decline to approve the move. Without cause means without alleging misconduct. The committee can then send the decision to Anthropic’s employees for a vote.

OpenAI offers the clearest precedent for employee power in AI governance. During OpenAI’s so-called “blip,” its board briefly fired CEO Sam Altman over concerns that he was not candid with them. Employee loyalty led the board to back down and reinstate him, and the company installed a new governance structure.

At Anthropic, the employee vote appears to formalize that kind of power. If employees back Anthropic’s board over the LTBT on a contested seat, they can overrule the trust. The reverse is also possible: a break among the founders could give common shareholders more say than expected.

Employees become a tie-breaker when the board challenges the trust over a director.

When do the founders and trust lose control?

The founders’ control has a defined endpoint. Ten years after the IPO, they can choose to end it. Their control also ends automatically if fewer than three of the seven founders remain.

Once founder control ends, the LTBT stays in place until 85% of common shareholders vote to end it. The trust can also step down with the board’s approval. If and when neither special class of shareholder exists, common shareholders finally gain control.

The thresholds make any transfer slow and deliberate. Common shareholders cannot remove the LTBT by ordinary majority; they need 85%. The founders do not have to wait for that vote to surrender their own voting block, although their automatic trigger arrives when their number falls below three.

Ordinary shareholders get the final say only after both special voting systems disappear.

What should investors watch next?

Anthropic’s structure creates three possible centers of power: the founders, the LTBT and the corporate board. Employees can break a board-versus-trust dispute. Common shareholders gain leverage if the founders split, or after the long-dated controls expire.

The market stakes are direct. Public money will back a frontier AI company, but shareholders will have little say over board composition or major safety decisions. The policy stakes are equally grounded: Anthropic has made safety part of its corporate purpose, yet shareholder voting cannot readily redirect that purpose.

The hardest prediction is not what the governance documents say, but how the relationships work in practice. The first signals will be whether the seven founders remain aligned, whether the board invokes the employee vote, and whether anyone moves to shorten the ten-year path.

The arrangement’s real test will come when the founders disagree or the board challenges the trust, because those moments will show whether the IPO’s employee and public-shareholder checks work in practice.