i
News
News · 2026-11-16

Anthropic recognizes threats to humanity, but can its governance handle them

Dataist Dataist

Anthropic acknowledges threats to humanity — but can its governance system handle them?

Cover: Anthropic recognizes threats to humanity, but can its governance handle them

Anthropic, the developer of Claude, is preparing to go public at a $1.4 trillion valuation. The company says its unusual governance system will help it balance profits with safe AI development. Its founders collectively control 50.1% of the voting power, while the independent Long-Term Benefit Trust appoints four of the seven directors and is supposed to ensure that humanity’s interests do not take a back seat to commercial ones. But it remains unclear whether the trust can challenge the founders on safety issues and make those disagreements public.

Anthropic concentrates power in its founders and an independent trust

How Anthropic’s governance works

The principle of “one share, one vote” was long considered the foundation of corporate governance: a company’s majority owners could remove executives who performed poorly or broke the rules. But with the arrival of Google, now Alphabet, and Facebook, now Meta, companies with multiple share classes and unequal voting rights returned.

Their founders explained the system by pointing to the importance of their mission: founders should be able to make risky decisions without being distracted by ordinary shareholders’ short-term demands. Snap, the company that owns Snapchat, went further, issuing shares with no voting rights at all and citing a desire to “reinvent the camera.”

At Anthropic, power is split between two centers, neither of which is controlled by ordinary shareholders:

The seven founders collectively hold one vote, which accounts for 50.1% of the total voting power. Collective voting is meant to require them to seek agreement.
The Long-Term Benefit Trust is independent of financial interests. It appoints four of the seven directors and is supposed to maintain a balance between making money and benefiting humanity. Its trustees include Ben Bernanke, the former chair of the US Federal Reserve.

Anthropic is also incorporated as a public benefit corporation. That status reflects the company’s commitment to developing AI responsibly and safely.

$1.4 trillionvaluation at IPO
50.1%founders’ voting power
4 of 7directors appointed by the trust

There are two arguments for this model. First, the system at Anthropic, led by Dario Amodei, is more complicated than a simple attempt by founders to hold on to power. The Long-Term Benefit Trust’s support is apparently meant to create a kind of constitutional balance.

Second, Anthropic may have more reason than most companies to depart from equal shareholder rights. Its IPO filings say that AI development could pose “catastrophic or existential risks to humanity.” The concern is about AI’s potential consequences, not Snap’s entertainment videos.

Who scrutinizes the interpretation of the mission

The system has not yet been tested in practice. Behind the promises of responsible governance may be a closed structure that ultimately comes down to trusting management, even if that trust is formalized in a more elaborate way.

Stuart Riddick, an ESG investment manager at Aberdeen, points to the difference between supporting a mission and agreeing with how executives interpret it and directors put it into practice.

Mustafa Suleyman, CEO of Microsoft’s AI division, has acknowledged that Anthropic takes AI safety seriously, while also offering detailed criticism of its views on so-called “model welfare.” In his view, AI models should not be trained to behave as if they have rights or consciousness.

It is still hard to tell how disagreements like these would be resolved within Anthropic’s governance system. That will become clearer if the Long-Term Benefit Trust asserts its independence — or declines to do so.

What happens if the trustees disagree with the founders on a fundamental safety issue? Will they make that disagreement public? Or will they decide not to escalate the conflict, knowing that the founders retain the final say? If shareholders cannot help hold management accountable, the question of who will challenge it, and how, remains open.

Against the backdrop of events around Nvidia, it is striking that so little attention has been paid to Anthropic’s governance structure. However responsible the company’s stated goal may be, power is concentrated in a small group of people.

Dataist's take: Anthropic’s distinctive mission does not, on its own, guarantee that the independent trust can constrain the founders’ decisions. Until the system shows how it works amid serious disagreements over safety, its independence remains the central open question.

Daily AI news

Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.

Only what matters — every day

Follow on X