Anthropic's Super-Voting Shares: A Governance Paradox for the Decentralization Ethos

We audit the code, but who audits the conscience? When Anthropic, the AI safety lab built on a promise of responsible development, signals a move toward super-voting shares for its CEO and co-founders ahead of an IPO, the question becomes painfully acute. This isn't just a corporate governance story; it's a values conflict dressed in the language of stock classes. For those of us who have spent years in the trenches of blockchain governance—watching DAOs struggle with plutocracy, analyzing the concentration of hash power in Bitcoin mining pools, and auditing the ethical implications of smart contracts—this move feels like a familiar, unsettling echo. Anthropic, founded by former OpenAI researchers, has long positioned itself as the counterweight to the 'move fast and break things' ethos. Its mission is explicitly anchored in AI safety and long-termism. The announcement that it plans to issue super-voting shares to its leadership, giving them disproportionate control over strategic decisions, is ostensibly a mechanism to protect that mission from short-term market pressures. Yet, as we've learned in the blockchain world, the concentration of power—even with the best intentions—can become a Trojan horse for centralization, vulnerability, and the erosion of the very values it seeks to protect. The road to hell is paved with good intentions, and the architecture of control is its foundation. From a technical perspective, the super-voting structure is a classic dual-class share arrangement. Typically, the founders get shares with 10x or 20x voting power, while public investors receive ordinary shares with one vote each. This is not new—Google, Meta, and Snap have all used it. But in the context of an AI lab that claims to be building for humanity, the stakes are higher. The core question is not whether the founders deserve control, but whether the system provides sufficient checks and balances to ensure that control is exercised in the public interest, not just the founders' vision. Based on my experience auditing governance models in early DAO projects—like the 1Balance project where I identified three critical voting centralization risks in 2017—I’ve seen how seemingly benign mechanisms can harden into unaccountable power structures. In a blockchain context, we talk about 'trust minimization' and 'code is law.' But when the code is controlled by a handful of individuals, the law becomes their interpretation. The same applies to Anthropic. The super-voting shares act as a governance lock, preventing the board or shareholders from overriding the founders' decisions. Theoretically, this could protect against a hostile takeover that prioritizes profit over safety. Practically, it also means that if the founders' judgment on safety becomes flawed—or if they conflate safety with their own authority—there is no external mechanism to correct course. Here, the contrarian angle emerges. In the crypto world, we often vilify centralization, but we also recognize that some forms of concentration are necessary for efficiency. Bitcoin's mining pools are a classic example: despite the ideal of decentralization, power has coalesced into three major pools. We accept this because the underlying protocol enforces consensus rules, and the economic incentives of miners are aligned with the network's health. Similarly, Anthropic's super-voting shares could be seen as a 'mining pool' for governance—a necessary concentration to ensure consistent execution of a long-term vision that spans decades. The key difference is that Bitcoin's pool concentration is transparent and contestable; miners can switch pools. Anthropic's governance concentration is opaque and permanent, unless sunset clauses or other mechanisms are introduced. But there is a deeper, more uncomfortable truth. The argument that super-voting shares protect AI safety is intellectually appealing, but it creates a dangerous precedent. It assumes that the founders' definition of 'safety' is correct and static. History shows that power corrupts, and even the most principled leaders can become insulated from feedback. In the blockchain space, we have seen countless DAOs fail because they lacked accountability mechanisms for their founders or core teams. The DAO hack of 2016 was a direct result of a governance failure—the code was exploited because the human oversight was insufficient. If Anthropic's founders make a catastrophic error in judgment—say, an overconfident deployment of a powerful AI model—the super-voting structure would make it nearly impossible for the board or investors to intervene. The very mechanism designed to protect the mission could become the instrument of its undoing. From an investment and valuation standpoint, the market will likely price in a governance discount. Studies by ISS and Glass Lewis consistently show that dual-class companies trade at a 5-10% discount relative to single-class peers. For a company like Anthropic, which raised billions at high valuations, this discount may be tolerable, but it creates friction. It also raises questions about the role of institutional investors. Vanguard, BlackRock, and other large asset managers have increasingly pushed for sunset provisions and independent board oversight. If Anthropic's IPO proceeds without such safeguards, it may face a cooler reception from index funds and ESG mandates. The company's high scarcity value—as one of the leading AI labs—may offset this, but the tension is real. Yet, the most significant risk is not valuation but trust. The crypto community has learned that trust is earned in silence, lost in noise. If Anthropic's governance becomes a symbol of 'safety theater'—a structure that looks protective but is actually a power grab—it will erode the very trust that the AI safety movement relies on. The public already struggles to differentiate between genuine safety efforts and marketing. A super-voting arrangement, without transparent checks, risks being seen as a shield for unaccountable authority. What then is the path forward? I believe Anthropic could adopt a hybrid model that draws from blockchain governance best practices. For example, they could implement a 'safety sunset clause' that converts super-voting shares to ordinary shares if independent safety audits fail or if a catastrophic incident occurs. They could establish an independent ethics committee with veto power over certain high-risk decisions, similar to a DAO's multi-sig treasury. They could also commit to publishing a 'governance whitepaper' that details the rationale for the structure, the voting power ratios, and the safeguards in place. This would not only satisfy institutional investors but also align with the company's stated values of transparency and responsibility. Build not for the peak, but for the plain. The peaks of hype and control are tempting, but the plain—the steady, accountable, and resilient governance—is where long-term trust lives. Anthropic has a unique opportunity to set a new standard for AI governance, one that balances founder vision with decentralized accountability. The super-voting shares are not inherently wrong, but they must be paired with mechanisms that ensure the conscience remains auditable. Otherwise, we are left with a code that only its creators can read, and a conscience that only they can judge. In a world where AI will increasingly shape our lives, that is a risk we cannot afford to take.