The Long-Term Benefit Trust is a cage, not a shield.
Last week, news broke that Anthropic is structuring its IPO with a novel governance mechanism: a Long-Term Benefit Trust (LTBT). The stated goal is to ensure its AI models remain aligned with human values, long after the founders have stepped away. It sounds noble. It reads like a permanent commitment to safety.
But as someone who has spent the last decade watching governance mechanisms fail in both traditional finance and decentralized protocols, I see a different pattern. The protocol is not the promise. The people holding the keys are.
Context: The Anthropic-SpaceX Playbook, Diverged
To understand the LTBT, we need to revisit Elon Musk’s IPO playbook for SpaceX. He famously refused to take SpaceX public for years, citing the tyranny of quarterly earnings and hostile activist investors. When he finally filed, he created a dual-class stock structure that gave him super-voting rights. The logic was clear: protect the long-term vision from short-term market pressure.
Anthropic’s LTBT is a variant of that. Instead of concentrating control in a single founder, it places ultimate authority over model alignment—not profit—into a legally binding trust. The trust’s board would be composed of ethicists, scientists, and regulators, not shareholders. The idea is that even if a future CEO wanted to prioritize growth over safety, the trust could veto.
Based on my experience auditing governance failures in both DeFi and traditional asset management, the mechanism sounds elegant in theory. But in practice, it introduces a paradox that no one in the article seems to be addressing.
Core: The Governance Paradox of Immutable Intent
The core insight is that the LTBT is attempting to solve a problem that cannot be solved by immutability alone. It replicates the flawed logic of a smart contract without a kill switch, or a DAO without a quorum.
Here is the issue: the trust’s power is absolute, but its members are human. They will be chosen by a process that is opaque by design. They will be subject to internal politics, regulatory capture, and evolving social norms. The trust’s charter says “long-term benefit,” but who defines that term? In 2025, the LTBT board might interpret it as “don’t release a model that can generate autonomous weapons.” In 2035, a new board might interpret it as “don’t release a model that threatens the economic stability of the nation.”
During the Terra/Luna trauma of 2022, I watched a similar governance failure unfold. The Anchor Protocol was designed to be a decentralized savings account, but its governance was effectively controlled by a small group of early insiders with veto power. They promised safety, they promised stability. When the protocol needed to adapt, the governance structure froze. The system was too rigid to respond to a changing market. The protocol held, but the consensus fractured.
Anthropic’s LTBT faces the same risk. It is a centralization of ultimate authority disguised as a safeguard. The trust’s power is not derived from code, but from a piece of paper. And paper can be rewritten, or ignored, when the will is strong enough.
Contrarian: The Decoupling Thesis is a Mirage
The article positions the LTBT as a major differentiator from SpaceX’s model. But the contrarian angle is that the two structures are converging, not diverging. Both are attempts to decouple control from capital. SpaceX’s dual-class stock decouples voting power from economic ownership. Anthropic’s LTBT decouples moral authority from both. In both cases, the stated goal is long-termism.
But here is the blind spot: every major governance innovation in tech history has eventually been captured by the very forces it was designed to resist. The original founders of Google created a dual-class structure to protect their “Don’t Be Evil” ethos. By 2025, the same company was cutting ethical AI staff and serving ads to genocidal regimes. The structure didn’t protect the ethos; it protected the founders’ ability to change the ethos.
In the deep end, liquidity is the only oxygen. And the ultimate liquidity is the ability to change the rules. The LTBT, by its very nature, makes that harder. But it does not make it impossible. It just creates a higher barrier to entry for the first change. Once that barrier is broken, the entire structure collapses.
Takeaway: The Cycle of Trust, Broken Again
I am not saying the LTBT is a bad idea. I am saying it is a dangerous one if we mistake it for a permanent solution. It is a cage, but cages can be unlocked. The real question is not whether the mechanism works, but whether the people inside it can be trusted to stay in their cage.
Pattern recognition is the only true hedge. The pattern here is clear: every time a governance innovation is presented as a final answer, it is a signal that the next crisis is already being engineered. The LTBT will be tested. Maybe not in 2025, but by 2030. When it is, the world will ask: who really controls the keys to the future?
Alpha is not found; it is harvested from chaos. And the chaos is coming.
Art was the asset, but attention was the currency. The LTBT is not an asset; it is a promise. And promises are only worth what the market believes them to be worth.