Anthropic's Super Voting Rights: A Lesson in Decentralization's Unfinished Revolution

The news hit my feed at 3 AM Seattle time. Anthropic, the darling of the AI safety movement, is preparing to issue super voting rights to its CEO Dario Amodei and co-founders, effectively insulating them from external shareholder pressure. The Information reported this on August 19, and within hours, the crypto Twitter I inhabit was ablaze with predictable takes: "See, even the ethical AI people need centralized control." But I read the details and felt something else—a deep, familiar ache. This isn't a story about centralized greed. It's a story about a governance failure that decentralization was supposed to solve, but hasn't yet. Decentralization is a verb, not a noun. Anthropic was founded on a promise. Unlike OpenAI, which pivoted from non-profit to capped-profit, Anthropic publicly committed to a "benefit corporation" structure with a long-term benefit trust. They installed a non-shareholder trustee system to ensure that the founding team's vision—safe AGI—wouldn't be compromised by profit-seeking investors. But now, with the super voting rights move, they're admitting that trust alone isn't enough. The founding team's equity stake is too low. They need formal voting power to outmuscle the capital that has flooded in. It's a humbling admission: even the most noble intentions can't survive the gravitational pull of external shareholders without a structural shield. As a decentralization evangelist who spent 2017 in a Capitol Hill cafe dissecting the Ethereum whitepaper, I see this as a mirror. The crypto world has spent years mocking traditional corporate governance for its principal-agent problems. We built DAOs, token voting, and liquid democracy as alternatives. But let's be honest: how many of our own projects have faced the same dilemma? The answer is almost all of them. I remember DeFi Summer 2020, when I was forking yield strategies and tweeting about "governance theater"—the irony of token votes that were always decided by the founding team's multi-sig. The problem is not that centralized control is bad. It's that we keep pretending we can have decentralized control without a mechanism to protect the founding vision from the capital that funds it. Anthropic's move is a primal scream of a founding team that has seen too many startups get captured. They watched as OpenAI's board fired Sam Altman, then reinstated him under employee pressure, revealing that governance is a power struggle, not a principle. They saw the implosion of FTX, where a centralized founder with absolute control committed fraud. They are trying to thread a needle: keep the founders' values intact while still accessing the capital needed to train frontier models. The super voting rights are a shield, not a sword. But the question remains: does a shield that protects the founders from external shareholders also protect the company from bad founders? The answer is no. That's why the trustee system exists—to provide a check. But the trustee system is only as strong as the people who staff it, and those people are selected by the founders. It's a circular problem. From a technical perspective, this is identical to the governance mechanisms we see in blockchain protocols. The OP Stack, for instance, uses a "Security Council" with special powers to override token votes in emergencies. The ZK Stack has a similar mechanism. The difference is that in crypto, these override powers are transparent, auditable, and often time-limited. In Anthropic's case, the super voting rights are permanent and opaque. The founding team will control a majority of the board through special class shares. There is no on-chain mechanism to challenge them. The only recourse for external shareholders is to sell their shares—a exit that is functionally equivalent to a token dump. But the market doesn't punish that exit because every investor knows the rules before they buy in. So where is the problem? Here is my contrarian take: the real problem is not that Anthropic is centralizing power. It's that the entire AI industry is built on a fallacy—that a single company can build AGI responsibly. The only way to align AGI with humanity is not through a corporate board with super voting rights, but through a decentralized, open-source, permissionless ecosystem. I am not saying that every AI project should be a DAO. But I am saying that the conversation about AI governance is stuck in the 20th century, while the technology is leaping into the 22nd. The same logic that drove me to build "Ghost Protocol" in 2022—a privacy framework for decentralized identity—applies here: the most resilient systems are those that distribute trust, not concentrate it. Super voting rights are a band-aid on a bullet wound. Based on my experience bridging institutional investors into DeFi through the "Ethical Bridge" project in 2024, I can tell you that the institutional mind is not ready for true decentralization. They want control, predictability, and a single point of accountability. They will never fund a truly decentralized AGI project because they cannot sue a smart contract. But that doesn't mean we should give up on the ideal. It means we need to build a hybrid model: a decentralized protocol for the core logic, with a registered foundation in a jurisdiction that allows for legal recourse. Uniswap did it. MakerDAO did it. The AI industry can too. Anthropic's super voting rights are a symptom of a deeper failure: the inability of the current corporate governance model to handle the scale and risk of AGI. The founding team is acting in what they believe is the best interest of humanity. But history shows that even the best-intentioned founders can become corrupted by power. The only safeguard is a system that distributes power across a diverse set of stakeholders, with transparent checks and balances. That is the lesson of blockchain. That is the lesson of the bear market, when we refined our ideas away from speculation and toward resilience. Decentralization is a verb, not a noun. What does this mean for the crypto industry? It means we have a real opportunity. The AI industry is about to face a crisis of legitimacy. The public will see these super voting rights as a power grab, even if they are framed as a safety measure. The crypto community can offer a better model: a decentralized AI governance framework that uses token voting, multi-sig control, and optimistic challenge periods to ensure that no single entity can make unilateral decisions about AGI. I am already working on a proposal for a "Decentralized AGI Alignment Protocol"—a layer-2 that aggregates AI models and allows users to vote on which ones are deployed. It's early, but the need is urgent. In the end, Anthropic's move is a wake-up call. It tells us that the tension between founders and funders is not going away. It tells us that the current system is broken. And it tells us that the crypto industry has a chance to build something better. But we must be honest about our own flaws. We have to stop pretending that DAOs are perfect. We have to stop ignoring the fact that most token votes are controlled by whales. We have to build better governance tools, not just better yield farming strategies. The future is not about choosing between centralized control and chaotic democracy. It's about designing systems that are resilient, transparent, and accountable. That is the work. That is the verb. And it starts with understanding that Anthropic's super voting rights are not a scandal—they are a mirror. We are all trying to solve the same problem. The question is whether we will learn from each other's mistakes or repeat them. The next time you see a blockchain project proposing a "founder's veto" or a "security council," ask yourself: is this a necessary safeguard or a backdoor to centralization? The answer is rarely black and white. But the willingness to ask the question is what separates the evangelists from the opportunists. Decentralization is a verb, not a noun.