Anthropic's Super-Voting Rights: A Governance Audit from the Crypto Trenches

The code is a hypothesis waiting to break. Last week, The Information reported that Anthropic—the AI lab behind Claude—is preparing a new class of stock with enhanced voting rights for CEO Dario Amodei and co-founders. The stated goal: insulate the founding team from external shareholder pressure. On the surface, this is a typical Silicon Valley power play. But to a Layer2 researcher who has spent years dissecting DAO governance, smart contract voting modules, and the entropy of decentralized decision-making, this move reveals something deeper. It is a confession that the governance model of even the most cutting-edge AI company is structurally fragile. And it mirrors, in eerie detail, the governance failures I have seen in crypto protocols pretending to be decentralized. Context: Anthropic’s current ownership structure is scattered. Co-founders hold a relatively low percentage of equity. The investors—from FTX’s estate to Google—hold significant sway. The new super-voting shares will allow the founding team to elect a majority of board members, even with minimal economic stake. They also plan to retain a non-shareholder trustee system, a layer of insulation that further reduces external influence. This is not novel. Google, Facebook, and Snap all have multi-class structures. But Anthropic is different. It positions itself as a "public benefit corporation" with a mission to build safe AI. Super-voting rights directly contradict the transparency and accountability that safety demands. The trust assumption is that the founders will always act in the best interest of humanity. That assumption is a hypothesis waiting to break. Core: Let me trace the architectural trade-offs. In crypto, governance tokens are the equivalent of voting shares. I have audited protocols where the founding team held 10% of the tokens but 90% of the voting power through a governance module with quadratic voting or time-weighted multipliers. The result is always the same: the protocol becomes a reflection of the founders’ will, not the community’s. The code enforces a centralized hierarchy. Anthropic’s super-voting shares are the same pattern, but with legal contracts instead of smart contracts. The gas leak in this untested edge case is the assumption that the founders will remain aligned with the mission. In crypto, we have seen this fail repeatedly. The DAO that was "too big to fail" collapsed because the founding team’s incentives drifted toward liquidity extraction. The same will happen at Anthropic if the founders decide—rationally or not—that short-term profit or personal influence outweighs safety. Modularity isn’t an entropy constraint. Anthropic’s move is an attempt to modularize governance: separate control from ownership. But modularity without clear boundaries creates brittleness. The super-voting class is a single point of failure. If the founders’ private keys (metaphorically) are compromised, the entire governance structure collapses. In crypto, we mitigate this with multisig, timelocks, and progressive decentralization. Anthropic’s approach is the opposite: it centralizes control into a small set of human actors. This is fine until one of those actors makes a mistake. And humans make mistakes. I have seen this pattern in the cross-chain bridge I audited in 2025. The optimistic verification module had a single point of trust in the acting validator set. The code was elegant, but the governance was fragile. The same design flaw applies here. Let me get specific. The super-voting shares will allow the founding team to elect a majority of board members. This means the board becomes a rubber stamp for the founders. In crypto, we call this "the illusion of decentralization." The board’s fiduciary duty to shareholders is subverted by the voting structure. The non-shareholder trustee system adds another layer of opacity. Who are these trustees? How are they chosen? The code is silent. In a smart contract, I would flag this as a permissioned upgrade mechanism. It is a backdoor. The same logic applies to Anthropic’s governance. The lack of transparency in the trustee selection is a governance vulnerability that could be exploited by a sophisticated actor. The bull market euphoria around AI is masking these structural flaws. Everyone is focused on the model’s capabilities, not the governance that will decide how the model is deployed. Contrarian: The conventional wisdom is that super-voting rights protect long-term vision against short-term shareholder pressure. That is true in theory. But the blind spot is that the founders themselves are subject to short-term pressures. The pressure to ship the next model, to raise the next round, to compete with OpenAI. The super-voting structure does not eliminate that pressure; it only removes the formal check. In crypto, we call this "the dictator’s dilemma." The dictator has absolute power but cannot be replaced, so any mistake becomes catastrophic. Anthropic’s model is the same. If Dario Amodei makes a bad strategic decision, there is no governance mechanism to correct it. The board is controlled by him. The external shareholders have no say. The only exit is to sell the stock, which further concentrates power. This is the opposite of resilience. It is a brittle system that relies on the assumption that the founders are infallible. Based on my experience auditing DAO governance, I can tell you that infallibility is a myth. The best protocols have built-in checks, like veto councils or on-chain voting with quadratic weighting. Anthropic has none of this. Another blind spot: regulatory risk. Regulators are increasingly scrutinizing super-voting structures, especially in companies with public benefit charters. The SEC has already signaled that such structures can violate shareholder rights. In crypto, we see the same tension: the SEC’s actions against protocols that claim to be decentralized but have centralized control. Anthropic’s move puts a target on its back. The company’s narrative of "safe AI" becomes harder to maintain when the governance is opaque and concentrated. The institutional investors who funded the last round may not be happy. They are now locked into a structure where they have no influence. The code is a hypothesis, but the legal consequences are very real. Takeaway: Anthropic’s super-voting rights are a governance patch, not a solution. They mask the underlying problem: the founders do not want to give up control, but they also want outside capital. This is the same tension that drives the creation of centralized sequencers in Layer2 chains. The sequencer is the control point, and the token holders are the shareholders. The solution in crypto is to progressively decentralize the sequencer, starting with a permissioned set and moving to a decentralized set over time. Anthropic could do the same: issue super-voting shares that sunset after a certain period, or create a governance council with independent members. But they haven’t. The result is a governance model that is optimized for the founders’ comfort, not for long-term resilience. The question every reader should ask is: when the hypothesis breaks, who will fix it?