Anthropic's dual-power governance model faces scrutiny
The AI firm's unique structure balances founder control with a non-profit trust to manage existential risks, but critics question its actual oversight.

The short version
- The seven founders collectively hold 50.1% of the ultimate voting power, requiring them to reach a consensus.
- A Long-Term Benefit Trust nominates four of the seven directors to balance profit with human benefit.
- The company is structured as a public benefit corporation to signal a commitment to responsible AI development.
- Critics worry the structure may be an inward-looking 'trust us' model that lacks independent accountability.
Why it matters
As AI becomes more powerful, the way these companies are governed determines whether safety remains a priority or becomes secondary to growth. This case highlights the tension between corporate mission statements and the reality of concentrated executive power.
A departure from standard tech
Most large technology companies operate on a principle where ownership corresponds to voting rights. In this traditional model, shareholders typically have the power to remove underperforming managers or influence the company's direction through a simple majority.
However, many modern tech giants have moved toward dual-class structures to protect the vision of their founders. These arrangements allow creators to pursue long-term goals without being distracted by the immediate demands of ordinary investors.
Anthropic is taking this concept further by creating a governance system that ignores the standard 'one share, one vote' rule. As reported by The Guardian, the company is moving toward a $1.4tn listing with a structure that places power in two specific centers rather than with the shareholders.
The dual-power structure
The first center of power belongs to the seven founders of Anthropic. They collectively hold 50.1% of the ultimate voting power, a setup specifically designed to force the group to reach a consensus on major decisions.
The second center of power is the Long-Term Benefit Trust. This is a financially disinterested entity that nominates four of the seven directors to ensure the company balances its drive for profit with the broader benefits to humanity.
The trust includes notable figures such as Ben Bernanke, who previously chaired the US Federal Reserve. By using this structure, Anthropic aims to create a constitutional balance of power that prioritizes the responsible development of artificial intelligence.
Justifying the deviation
Anthropic argues that its unconventional governance is necessary because of the unique risks posed by AI. The company's listing documents state that the development of these technologies could pose catastrophic or existential risks to humanity.
Because of these high stakes, the firm believes it requires more than the standard corporate oversight used by social media companies. The goal is to move away from a simple power grab by founders and toward a model that accounts for global safety.
The company is also designated as a public benefit corporation. This legal status serves as a formal nod to its commitment to developing AI in a way that is both safe and responsible for the public good.
Questions of independent oversight
Despite these high-minded goals, some observers worry that the governance model is essentially an elaborate way to maintain founder control. Critics suggest that the setup might simply be an inward-looking system where the company asks the public to trust them without external checks.
Investment managers have noted that agreeing with a mission in principle does not mean everyone agrees on how that mission is actually implemented. For example, different leaders in the industry have expressed strong disagreements over specific concepts like model welfare.
It remains unclear how Anthropic will handle internal conflicts between the trust and the founders. If the trustees disagree with a founder on a safety question, it is unknown whether they will speak out or yield to the fact that the founders hold the ultimate vote.
The concentration of power
The current arrangement represents an extraordinary concentration of power in the hands of a very small group of individuals. While the trust provides a layer of oversight, the final say on the company's direction remains with the founding team.
If shareholders are denied any role in holding executives to account, the true level of challenge to the leadership is difficult to measure. The effectiveness of the Long-Term Benefit Trust depends entirely on its willingness to act independently of the founders.
As Anthropic moves forward with its listing, the world will be watching to see if this experiment in governance actually works. The company must prove that its structure can successfully navigate the tension between commercial success and the prevention of existential risk.
What to watch next
Observers will be looking for the first major instance where the Long-Term Benefit Trust must weigh in on a controversial safety decision. It will be important to see if the trust has the teeth to challenge the founders' primary voting power.
Additionally, the industry will watch how Anthropic handles specific technical disagreements, such as those regarding the rights or consciousness of AI models. These debates will test whether the company's governance is a functional safety mechanism or a symbolic gesture.
Try it
Look up the definition of a 'public benefit corporation' to see how it differs from a standard corporation in terms of legal obligations.
Watch
Will the Long-Term Benefit Trust ever publicly disagree with the founders?
How will Anthropic resolve conflicting views on 'model welfare' within its governance structure?
Will other AI firms adopt similar trust-based models to manage existential risk?
Sources and links
- ReportingThe Guardian: the article we sourced
First reported by The Guardian. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.