Anthropic's Super-Voting Shares: Governance Centralization Dressed as Mission Protection

The logic held until the ledger lied. Anthropic, the AI lab behind Claude, announces plans to grant super-voting shares to its CEO and co-founders ahead of an IPO. On the surface, this is a standard Silicon Valley move—locking founder control to ward off activist investors. But for anyone who has watched governance structures fail in crypto, this is a red flag dressed in a safety narrative. The same pattern that allowed DAO whales to extract value, that turned multi-sig committees into single points of failure, is now being wired into the core of an AI company that claims to prioritize long-term risk mitigation. Context is simple: Anthropic has raised billions at a valuation exceeding $30 billion. Its mission is to build safe, beneficial AI. The IPO is the next logical step for liquidity and scale. Yet instead of embracing the transparency and accountability that public markets demand, the founders are erecting a wall. Super-voting shares—typically 10x or 20x voting power—mean that ordinary shareholders, including institutional investors, will have little say in strategic direction. The argument is that this protects the mission from short-term profit pressures. It is the same argument that Meta and Google made, but for an AI company whose product could shape society, the stakes are higher. Core analysis: this is a governance attack vector, plain and simple. From my years auditing smart contracts, I've learned that centralization is the root of all exploits. In 2020, I simulated a governance attack on Compound's cETH contract. I found a 12-second window where a flash loan could drain liquidity because the protocol lacked slippage protection. The team ignored it. The same structural blindness is present here. Super-voting shares concentrate decision-making power into a handful of individuals. There is no sunset clause mentioned, no external oversight committee with veto power. The founders are the multi-sig, and the seed is their shared vision. Immutability is a promise, not a feature. When the vision diverges from reality—when market pressure mounts or a safety trade-off is needed—the internal checks will fail because there are no external ones. Let's dig into the technical specifics based on what is public. The voting power multiplier is likely 10x or 20x, as is standard. This means a founder with 10% economic ownership could control 67% of the voting power. In a typical IPO, this structure would trigger a governance discount of 5-10% from institutional investors like BlackRock and Vanguard. But for Anthropic, the discount may be larger because the company's entire value proposition rests on the credibility of its safety mission. A governance structure that concentrates power without transparency is a direct contradiction to the principle of distributed alignment. Governance is just a slower attack vector. The attack is not a flash loan; it's a slow erosion of accountability. Furthermore, the timing is suspicious. The IPO is not imminent—it's a plan ahead of the event. This suggests the founders are preemptively locking in control before any market scrutiny. In crypto, we call this a rug pull in slow motion. The team is ensuring that even if the public markets demand quarterly earnings and short-term growth, the founders can ignore them. That sounds good for safety, but it also means that if the founders decide to pivot to profit-only AI, there is no mechanism to stop them. The same power that protects the mission can also protect the opposite. Consider the competitive landscape. OpenAI has a non-profit capped-profit structure, which is messy but includes a board that can remove the CEO. xAI is private with Elon Musk as sole visionary. Anthropic's super-voting model is a hybrid: public liquidity with private control. It's worse than both because it combines the lack of transparency of a private company with the capital access of a public one. The bulls will say this is necessary to avoid the fate of OpenAI's near-collapse over governance disputes. But the counter is that Anthropic's structure eliminates the very checks that could prevent a founder from steering the company off a cliff. Trace the hash, ignore the hype. The hash here is the governance structure, and the hype is the 'safety first' narrative. Contrarian angle: what the bulls got right. The super-voting structure does provide a buffer against short-termist hedge funds that might push for faster monetization of AI capabilities. Anthropic's mission is genuinely long-term, and the founders have a credible track record of prioritizing safety over profit. The ability to make rapid decisions without quarterly earnings pressure could allow them to invest in alignment research, slow down model releases, or call for regulation. In a world where AI companies are racing to deploy, this structure might be the only way to ensure that a board of directors doesn't fire the safety CEO. But this argument assumes that the founders are infallible and that their definition of 'safety' aligns with the public good. That is a fragile assumption. Takeaway: The chain remembers what you forget. Anthropic's governance structure is a promise, not a feature. When the IPO comes, the ledger will show who really holds the keys. Trace the power, ignore the hype. Every exploit is a history lesson in slow motion. This one is still being written, but the code is already in place. The question is not whether the founders have good intentions, but whether the system has enough friction to survive a hard fork of their own making.