The Czar and the Code: Jay Clayton, Self-Regulation, and the Decentralization Myth We Keep Telling Ourselves

We believe that systems should be governed by the people who live inside them. That conviction is why I traded a quantitative finance career for a rented apartment in Tallinn and a Discord server full of strangers who wanted to understand why "not your keys, not your coins" mattered more than any yield figure. It is why I have spent the better part of a decade explaining consensus mechanisms to people who will never read a yellow paper, using metaphors borrowed from village councils, neighborhood watches, and the awkward silences of family dinners. So when the rumor arrived on a gray Baltic morning — that Washington was preparing to hand the most consequential technology portfolio of our generation to a man whose entire professional life unfolded inside the machinery of securities law and Wall Street capital — I felt that particular vertigo that comes when the ground shifts beneath a belief you have organized your life around. Consider the moment when you realize that the people building the future do not share your definition of trust. The report, attributed entirely to unnamed sources, suggested that Jay Clayton — former chairman of the U.S. Securities and Exchange Commission, later nominated to serve as U.S. Attorney for the Southern District of New York — might be tapped as a new "AI czar." The details were thin. The framing was careless. In one breath the same report described him as the "US intel chief," a title that matches no public record of his career. That single error should stop any careful reader cold. It stopped me. Not because I care deeply about the org chart of the American executive branch, but because the error was a mirror. It reflected the same sloppiness, the same substitution of narrative for verification, that I have watched hollow out project after project in my own industry. The crypto world has spent a decade insisting that we can verify rather than trust, and yet we are just as susceptible to a confident headline as anyone else. Trust is the only currency that matters — and it devalues fast when the ledger is wrong. What is an "AI czar," and why should anyone outside Washington care? The word "czar" is American shorthand for a coordinator — a single figure appointed to cut across the jurisdictional knots of a sprawling bureaucracy. The United States has had drug czars, energy czars, Ebola czars. The label signals urgency and centralization at once: a problem so diffuse that no single agency can own it, and so important that someone must. Artificial intelligence fits that description precisely. The levers that shape AI's trajectory — compute infrastructure, electricity and data-center permitting, export controls, federal procurement, safety review, antitrust posture — are scattered across the Department of Commerce, the Bureau of Industry and Security, the Department of Energy, the Department of Defense, the White House Office of Science and Technology Policy, and a dozen independent agencies. A czar is an attempt to pull those threads into one hand. That attempt has a history. Under the previous administration, the governing framework was Executive Order 14110, which required developers of the largest frontier models — those trained above a threshold of roughly 10^26 floating-point operations — to report safety testing results to the federal government. It was, for all its imperfections, an attempt to install a check on the most powerful systems before they were deployed. Within days of taking office in January 2025, the new administration revoked it and replaced it with an "AI Action Plan" that framed regulation itself as the obstacle. The plan's language emphasized removing "burdensome" rules, accelerating deployment, and treating American AI leadership as a matter of national competitiveness. Into that vacuum steps the czar. The reported candidate, Clayton, spent his career in securities law and capital markets. He chaired the SEC from 2017 to 2020, a period when the agency's posture toward crypto was defined by enforcement rather than clarity — a posture that, I should say plainly, cost my own community years of ambiguity. He was a partner at Sullivan & Cromwell, one of the most powerful corporate law firms on earth. His instinct, his network, and his vocabulary all belong to the world of capital formation. That background is the signal. Personnel is policy. When you choose a securities lawyer to run an AI portfolio rather than a safety researcher or a systems engineer, you are announcing a hierarchy of values. You are saying that the primary challenge of artificial intelligence is economic — how to mobilize capital, accelerate deployment, and win a race — rather than epistemic or ethical. You are saying that the risks worth managing are the ones that show up on a balance sheet. And here is where my world intersects. Because the crypto industry has been living inside a version of this story for years. For those who have not watched the parallel unfold: the crypto ecosystem spent a decade promising that code would replace institutional trust. Smart contracts would enforce agreements without courts. DAOs would govern protocols without boards. The blockchain would be a ledger no one could quietly edit. And then, one by one, the seams showed. Upgrade keys sat in the hands of a handful of multi-sig signers. Foundation wallets held the tokens that decided votes. "Decentralized" governance turned out to be a small number of people wearing the costume of a crowd. Which brings me to the report's most telling phrase. The article described a world in which "self policing" technology CEOs would be trusted to manage AI's risks. Self-policing. I have heard that phrase before — from protocols that promised their communities would hold them accountable, right up until the moment a core team quietly changed the rules. The phrase "self-policing" is the most honest thing in the entire report, and it is also the most dangerous. Self-policing is the governance philosophy that says an industry, left alone, will discipline itself. It is the philosophy that animated the crypto dream, and it is the philosophy that has failed, repeatedly, in every domain where it has been tried at scale. Financial markets tried it before 1929 and again before 2008. Social media tried it through the 2010s and produced a decade of documented harms. And crypto tried it, in a thousand Discord servers, with the conviction of true believers. I know this failure intimately, because I helped build inside it. In 2017, at the height of the ICO boom, I audited more than fifty whitepapers using my training in financial engineering. I was looking for viable economic models, and I found twelve. Just twelve. The other thirty-eight were elaborate self-policing fantasies — projects that promised their token holders would govern wisely, that their treasuries would be managed prudently, that their founders would not dump their allocations. Almost none of them built the mechanisms to make any of that true. They did not need to. In a bull market, belief is the mechanism. That is the trap. Self-policing works precisely as long as it is not tested. It works when asset prices rise and everyone's incentives align with the narrative. It fails — catastrophically and all at once — the moment the incentives diverge. Code binds, but people break or build, and the code that is supposed to bind them is only as strong as the humans who hold the keys. The AI version of this trap is now being laid. If technology CEOs are trusted to police themselves, the question is not whether they will act responsibly in good times. The question is what happens when a model causes real harm — a medical misdiagnosis, a financial fraud executed at scale, a deepfake that moves an election, a bioengineering recipe that should never have been printed. Self-policing assumes the answer is "the company will do the right thing." History assumes otherwise. The multi-sig truth is the quiet engine of every governance scandal in crypto, and it is the exact mechanism that will govern AI if we let it. Here is what almost no one outside the industry understands. When a DAO claims to be decentralized, what it usually means is that token holders vote on proposals, and those votes are executed by a small group of administrators who hold the private keys to a multi-signature wallet. The multisig is the choke point. Three of five signers, or four of seven, can upgrade the smart contract, pause the protocol, or move the treasury. The community votes. The multisig decides whether to honor the vote. I have watched this dynamic play out so many times that I no longer find it scandalous. I find it structural. The original promise of DAOs was that smart contract upgrade rights — the power to change the rules of the game — would be distributed. In practice, upgrade rights are almost always retained by the founding team and their allies, because the alternative is terrifying: an immutable contract that cannot be fixed is a contract that cannot be rescued from its own bugs. So the founders keep the keys. The community gets the illusion of control. This is not a moral failing unique to crypto. It is the universal pattern of governance. Whoever holds the upgrade keys holds the real power, regardless of what the ballot box says. And the AI czar is, in this metaphor, a multisig signer with an unusually large share of the keys. He will not personally write the regulations or approve the models. He will coordinate the people who do. His power will be the power to decide which questions reach the table. The report never asked this question. It never asked whether the czar would have budget authority or personnel authority, or whether the role would be purely advisory — a coordinator with a title and no teeth. That omission matters more than the appointment itself. A czar without power is a press release. A czar with power is a new center of gravity. The compliance shield is the industry's most successful piece of theater, and the AI sector is about to discover how useful it can be. For years I have watched crypto projects perform decentralization as a legal strategy rather than a technical reality. The playbook is familiar. Incorporate a foundation in Zug or the Cayman Islands. Distribute tokens through a structure that looks like a community but functions like a cap table. Claim that the protocol is governed by no one. Meanwhile, the team wallets are traceable on-chain. The foundation's holdings are public. The "decentralized" treasury is controlled by the same people who wrote the whitepaper. None of this is hidden, if you know where to look. The blockchain is, after all, a transparency machine. The irony is that the very technology used to perform decentralization is the technology that exposes its absence. Every "anonymous" team has an on-chain footprint. Every "community" treasury has a signer set. The AI industry is entering its own compliance-shield era, and it is doing so with the government's blessing rather than its suspicion. When a czar is appointed to coordinate AI policy and the guiding philosophy is self-policing, the natural result is a structure in which companies write their own rules, document their own compliance, and report their own incidents — if they report at all. The shield is not a foundation in Zug. The shield is a voluntary framework with no enforcement mechanism. I am not naive about the alternatives. Heavy-handed regulation can entrench incumbents, because only the largest firms can afford the compliance departments. That is the strongest argument for the deregulatory turn, and it deserves to be taken seriously. But there is a difference between smart regulation that lowers barriers and no regulation that concentrates power. The self-policing route does not distribute power. It hands it to whoever already has the most. There is a structural parallel between the fragmentation of crypto's scaling layers and the concentration of AI governance, and it is not a comfortable one. I have written before about the Layer 2 problem, and I will return to it here because it illuminates something the AI debate keeps missing. When Ethereum's scaling roadmap produced dozens of Layer 2 networks, the industry celebrated. More chains, more capacity, more innovation. What actually happened was subtler and sadder. The same relatively small pool of users and liquidity was sliced into fragments. Bridges proliferated, each one a new attack surface. Capital that could have deepened a single ecosystem instead spread thin across a dozen competing ones. We did not scale. We subdivided. The AI governance conversation is at risk of the same error, inverted. Where crypto fragmented its scarce liquidity, AI governance is consolidating its scarce attention. The czar is an attempt to pull scattered authority into one place. And here is the tension: both moves — fragmentation and consolidation — can be described as progress, and both can quietly destroy the thing they claim to protect. The Layer 2 fragmentation destroyed user experience and liquidity depth. The AI governance consolidation threatens to destroy accountability. When power is spread across many agencies, no single person can be blamed, but no single person can act decisively. When power is concentrated in one coordinator, decisions get made — but the coordinator becomes a single point of failure, a bottleneck, and a target. The question is not whether centralization or decentralization is better in the abstract. The question is where the power actually sits, and whether anyone can see it. The convergence of AI and crypto is where this story becomes personal, and where the stakes stop being abstract. In 2025, I helped launch a research group — the Human-Centric AI Alliance — with fifteen experts drawn from ethics, law, and computer science. Our focus was narrow and urgent: how decentralized identity can protect privacy in an age of large language models. We synthesized findings from twenty academic papers and ten pilot projects, and we proposed a framework we called "Verifiable Human Interaction." The premise was simple. As AI systems become indistinguishable from humans in conversation, the ability to prove that you are a person — without surrendering your identity to a central authority — becomes a foundational right. The AI czar's portfolio touches this directly. If the guiding philosophy is self-policing, then the question of who verifies humanity will be answered by the same companies that build the AI. That is a conflict of interest so obvious it is almost invisible. The entity that profits from synthetic content is asked to certify which content is authentic. The fox is invited to design the henhouse and to file its own quarterly reports on how well the hens are doing. Decentralized identity offers an alternative: cryptographic proofs that establish personhood without centralized surveillance. But that alternative requires regulatory space — it requires a government willing to recognize that privacy-preserving verification is not a threat but a public good. A czar whose instinct is capital formation rather than civil liberties is unlikely to prioritize it. The window for building humane infrastructure is open now, and it will not stay open forever. The capital mobilization angle is the part the market will read most eagerly, and the part most likely to be misread. When a former SEC chairman is floated for a policy role, the market hears a specific melody. It hears IPOs. It hears institutional money. It hears a friendlier posture toward the companies that dominate the indices. Clayton's career was spent inside the machinery that decides who gets to raise capital and on what terms. In that world, the AI portfolio is not primarily a safety challenge. It is an industrial policy instrument. This is where I want to be careful, because the bull market wants a simple story and the simple story is rarely true. Yes, a deregulatory posture lowers compliance costs. Yes, it can accelerate deployment. Yes, it can improve the environment for AI-adjacent capital formation. But the same posture that accelerates deployment also accelerates the accumulation of risk, and risk that is not priced today is priced tomorrow — usually by people who did not choose to hold it. I watched this dynamic in crypto during the 2021 cycle. Capital flooded in. Valuations detached from fundamentals. The market rewarded narrative over mechanism. And when the cycle turned in 2022, the people who had been sold "self-policing" projects discovered that the policing had never happened. I spent that winter organizing weekly calls for three hundred community members — we called them Resilience Rounds — simply so people could share what they had lost and what they had learned. The churn in our community dropped by forty percent, not because we fixed anything, but because we told the truth. When I published "The Ethics of Failure," I argued that project collapses should be analyzed as human and systemic failures rather than as moral failures of the community. The same lens applies here. The most useful thing I can offer is not a prediction about the czar. It is a method for reading this kind of news. My method comes from the workshops. In 2020, when DeFi's complexity was driving newcomers to anxiety and loss, I founded TrustStack, a community initiative that hosted twenty live sessions explaining liquidity pools and impermanent loss to more than two thousand participants. We did not teach people how to maximize yield. We taught them how to recognize when a system was hiding its own center of gravity. The single most valuable question we gave them was this: where are the keys? Ask that question of any governance structure and the theater dissolves. In a DAO, the keys are in the multisig. In a foundation, the keys are in the foundation's wallet. In a self-policing industry, the keys are in the boardroom. And in a newly created czar's office, the keys are wherever the job description puts them — which is precisely the detail the report left blank. In 2021, I curated a project called Art for Access, minting five hundred free NFTs for underrepresented artists in Tallinn. I analyzed a thousand transactions to show how digital ownership could empower creators economically. The lesson that stayed with me was not about the technology. It was about who gets to decide. The artists who controlled their own keys told different stories than the artists whose work lived on someone else's platform. Ownership is not a slogan. It is a set of permissions, and the permissions are always held by someone. Here is the counter-intuitive angle, and it is one I have resisted because it cuts against the community I love. What if the AI czar is more honest than most DAOs? Consider the comparison without flinching. A czar is a public official. His name is known. His background is documented. His authority, once defined, is written into the machinery of government and subject to congressional oversight, judicial review, and the scrutiny of a free press. If he fails, he can be subpoenaed, sued, and voted out. The structure is centralized, yes — but it is legible. Now consider the average DAO. Its founders are pseudonymous or semi-anonymous. Its governance is a maze of proposals and delegate coalitions that few participants understand. Its upgrade keys are held by a multisig whose signers are rarely named. Its "community" is a Discord server that can be captured by a well-organized minority. If it fails, the founders face almost no consequence. The structure is decentralized in name — but it is opaque in practice. Which is more trustworthy: a visible center of power that can be held accountable, or a hidden center of power that pretends not to exist? This is the pragmatism test, and crypto fails it more often than we admit. We spent years criticizing institutions for their opacity while building systems whose actual governance was less transparent than a corporate board. We preached "code is law" while quietly retaining the right to change the code. The AI industry is about to run the same experiment at a larger scale, and the honest thing to do is admit that we have no moral high ground from which to lecture it. Culture eats blockchain for breakfast. The same is true of AI. The technology is never the whole story. The story is always about who holds the keys, who bears the risk, and who gets to look away. So where does this leave us? Not with a conclusion about Jay Clayton, whose appointment remains unconfirmed and whose actual powers remain undefined. The report's most important failure was not the "intel chief" error, embarrassing as it was. It was the assumption that the identity of the czar is the story. It is not. The story is the architecture of trust that a society builds around its most powerful technologies — and whether that architecture admits where its center of gravity lies. I keep coming back to the workshops, to the grandmothers learning about liquidity pools, to the artists learning that ownership is a set of permissions. What they taught me is that trust is not a technology. It is a relationship. It cannot be coded into a contract, policed into a framework, or delegated to a czar. It has to be earned, in public, with the books open. We are building the future, together — but only if we are honest about who is holding the keys while we build it. The czar will be appointed or he will not. The policy will shift or it will not. The only variable we control is whether we ask the question that the report forgot: not who is in charge, but who can see the levers, and who can hold them accountable when they are pulled. That question belongs to all of us. It always has.