On August 19, The Information reported that Anthropic is preparing a class of stock with enhanced voting rights for CEO Dario Amodei and other co-founders. This is the first time the AI safety startup’s leadership will hold super voting shares. Given that the co-founders own a relatively low percentage of equity, this move is framed as protecting the founding team's control from external shareholders. Anthropic also plans to retain its existing non-shareholder trustee system and use special class shares to elect a majority of board members, further reducing investor influence. At first glance, this is a corporate governance story. But for anyone who has audited smart contracts and DAO treasuries, it reads like a familiar playbook: centralization masked as protection, trust replaced by opaque voting mechanisms. Let me dismantle this from a crypto security perspective.
Context: The Hype Cycle of AI Governance
Anthropic is the poster child of AI safety, raising over $7 billion from investors including Google, Salesforce, and Spark Capital. Its mission is to build safe, interpretable AI. Yet here we are, watching the founding team engineer a governance structure that concentrates power, not distributes it. The industry is in a hype cycle where AI companies are valued by narrative, not by their code base or audit trails. Investors are pouring money into a black box, and the founders are now building a second black box for voting rights. Sound familiar? It should. In crypto, we call this a "multi-sig with a single signer" — a system that claims decentralization but vests ultimate control in a few keys. The trustee system is their version of a governance council, but without the on-chain transparency that even a flawed DAO provides.
Core: Systematic Teardown of the Governance Stack
Let’s treat this like a smart contract audit. We have three layers: the voting rights class, the trustee system, and the board election mechanism. Each is a potential vulnerability.
Layer 1: Super Voting Rights
Anthropic is issuing a special class of stock that gives co-founders enhanced voting power. This is the equivalent of a proof-of-stake system where the founding team holds 90% of the stake but only 10% of the financial risk. In crypto, we call this a "pre-mine" or "insider allocation." According to the report, Amodei and his co-founders hold a relatively low percentage of equity. By creating super voting shares, they are effectively leveraging a minority economic stake to control a majority of board decisions. This is a classic principal-agent problem dressed in legal clothing. In my audits of DAO governance, I have seen similar structures where a small group of whales (often early investors) can veto any proposal. The difference is that on-chain, you can see the votes. Here, the super voting rights are likely structured in a private agreement, invisible to the public. This is a red flag for anyone who values transparency. The corporate veil is thicker than any smart contract.
Layer 2: The Non-Shareholder Trustee System
Anthropic also plans to retain a non-shareholder trustee system. This is a mechanism where an independent trustee holds some decision-making power, but the trustee is not elected by shareholders. In crypto, this resembles a "multisig with a trusted third party" — a design choice that immediately triggers a security audit alarm. The trustee is intended to protect the mission, but who selects the trustee? How are they removed? What is the on-chain (or off-chain) record of their actions? Based on my experience auditing real-world asset tokenization projects, I have seen trustees become a single point of failure. They are often chosen by the founding team, creating a feedback loop of control. The lack of a public, auditable process for trustee selection is a vulnerability. If the trustee is compromised or captured, the entire governance structure collapses. This is not a hypothetical ; in 2022, I documented a case where a DeFi project's governance council was controlled by a single entity through a similar trustee loophole. The result was a $10 million rug pull.
Layer 3: Board Election via Special Class Shares
Finally, the special class shares will be used to elect a majority of the board. This means that even if external investors hold common shares, they cannot influence board composition. In crypto, this is the equivalent of a "founder-controlled mutlisig" where the keys are held by the same small group. The board becomes a monopoly of interests. The critical question: is there any mechanism for shareholders to remove a board member? In most DAOs, you can vote with your tokens. Here, token (share) ownership is disconnected from voting power. This is a structural flaw that undermines the very concept of stakeholder governance. The Information report notes that this arrangement is designed to "resist pressure from external shareholders." In other words, it is a firewall against accountability. When I audit a protocol, I flag any function that allows a single address to change contract parameters without a timelock or multisig. Anthropic's board election is that function, but applied to a corporate entity.
Contrarian: What the Bulls Got Right
Let me be fair. The bulls argue that Anthropic's mission requires long-term stability. AI safety is not a quarterly earnings game. The intense competition from OpenAI and Google means that the founding team must be able to make rapid, strategic decisions without being bogged down by short-term shareholder demands. They point to the fact that super voting rights have been used by companies like Google and Facebook (now Meta) to maintain founder control. In those cases, the companies outperformed market expectations. The counter-argument is that such structures allow for radical innovation. In crypto, the analogous argument is that Bitcoin's governance (through miners and core developers) is slow but stable. The bulls also highlight that the trustee system can act as a check on the founders, preventing them from acting purely in self-interest. In theory, a non-shareholder trustee could be a neutral arbiter. However, the lack of transparency on how the trustee is appointed makes this a theoretical rather than a practical safeguard. The bulls are right to worry about the volatility of a board that is constantly fighting with investors. But the solution they propose — centralization — is a band-aid on a broken system. The better approach would be to design a transparent, auditable governance structure that still allows for long-term vision, such as a time-locked shareholder vote with a supermajority threshold. This is something I have seen work in well-designed DAOs.
Takeaway: The Accountability Call
Anthropic's governance overhaul is a mirror for the crypto industry. We are watching a company that preaches safety and transparency choose opacity and control. This is not a scandal; it is a pattern. Every time a project says "we are protecting the mission," they are usually building a wall around their own power. The question for investors in both AI and crypto is the same: are you buying a technology or a narrative? The code is the truth. In this case, the code is a private contract, invisible to auditors. The takeaway is not to condemn Anthropic but to demand that any governance structure, whether in a DAO or a Delaware corporation, be auditable. Until then, trust is a liability. The next time someone pitches you a "decentralized" AI project, ask to see the voting rights contract. If it's hidden, it's a red flag. Anonymity is a feature, but centralization is a bug. And in this bull market for AI, the bugs are being hidden in plain sight. NFTs are art until you inspect the metadata hash. Corporate governance is a promise until you inspect the voting rights. I have seen this before. In 2021, I audited a DeFi project that claimed to be community-owned. The founding team held a special "governance key" that could override any vote. The project rug-pulled six months later. The code was the tell. Anthropic's code is not on-chain, but the pattern is identical. The only difference is the legal wrapper. The responsibility falls on investors to demand transparency. If the founding team is afraid of accountability, they are probably building something that cannot survive scrutiny. The future of AI governance should be open-source, not super-voting-share closed. The market is watching. The next move is either a step toward trust or a step toward the same old system.