A Weapon No One Issued
How great wealth became unaccountable power — and why artificial intelligence is where the last brake gives way
One of three connected essays. Read this and From the Pope to the End of the World (https://laudinum.substack.com/p/from-the-pope-to-the-end-of-the-world) first, then The Window and the Loop (https://laudinum.substack.com/p/the-window-and-the-loop). Responses welcome: perlmuttera@gmail.com (All three of these essays were written with AI assistance.)
My ideas, AI refinement. The argument, the framing, and the judgment in this piece are mine; an AI helped draft and refine the prose. Originally written June 2026. — Alex Perlmutter
I want to begin with what this is not. It is not an argument against wealth, or against capitalism, or against the people who build enormous things and grow rich doing it. I am a capitalist, and I mean that broadly: I do not think capitalism is merely an economic system. It is that, but it is also a reflection of human nature — the drive to accumulate and to build does not stop at the edge of a market; it shows up in every arena people operate in, and I do not think there is any part of human existence it does not touch. It is the engine behind nearly everything we have made, and nothing here is meant to slow that engine down.
The argument is narrower, and stranger. It is that beyond a certain scale, a fortune stops being a quantity of money and becomes a different kind of thing: a standing capacity to act on the world that needs no one’s permission. A weapon, in the sense that matters — not because anyone forged it as one or issued it, but because the capability simply accrues, and once it exists, the only thing standing between it and its use is the conscience of the person who holds it. That is the whole of it. The only brake is his own judgment. There is no statutory ceiling anywhere in the system, no permission required, and — as I will argue — increasingly no institution positioned to say no.
The occasion for thinking about this is Elon Musk, who in June 2026 became the first individual in history worth a trillion dollars after SpaceX went public. But this is not really about him — not his politics, his temperament, or the particular companies he owns. Strip the name out. He is simply the person standing, right now, in a position the structure itself created and will keep creating. If the danger were one unusual man, you could wait him out. Because the danger is the position and not the person, the thing worth thinking about is what the position consists of — and how little, at the top of it, holds it in check.
What a Fortune Can Mobilize
Start with the most concrete question: how much can one person actually bring to bear? Not net worth on paper — deployable, spendable power. Most of a fortune like this is illiquid stock; you cannot sell it all without collapsing its value, and selling means surrendering control of the companies that are the point. But you do not have to sell. You borrow against it.
After June’s offering, Musk’s stake in the newly public SpaceX — about 42 percent of the company — was worth on the order of $766 billion. None of it is spendable yet: newly public shares carry a lock-up, a standard agreement, commonly around six months, that bars an insider from selling or pledging them until it lifts. Once it does, the first move is to borrow against the stock rather than sell it, which keeps both the shares and the control they carry. But banks lend cautiously against a single concentrated holding, because a forced sale of millions of shares would crater the very collateral they are holding — so they advance only a fraction of its value, a figure called the loan-to-value ratio. For a position like this the realistic band runs from about 20 percent at the floor to perhaps 35 percent with aggressive hedging. The floor is not hypothetical: when Musk borrowed against Tesla stock to help fund the Twitter purchase, his banks set the advance at roughly 20 percent of the shares’ value. At that conservative rate the SpaceX stake alone yields about $150 billion in cash; structured harder, closer to $270 billion — without selling a single share.
He could go further still. Once the lock-up is past, over the following year he could sell down a slice and reinvest it into a broad, liquid portfolio. Diversified holdings frighten lenders far less than one concentrated stock, so they lend against them at much higher ratios — often half their value or more — and by spreading the loans across many banks, so that no single one hits its concentration limit, he could draw close to the full amount. The tax on selling would be only federal: he is domiciled in Texas, which levies no state income tax. Run all the way out, that path plausibly frees up something approaching $380 billion. I will not build the argument on that ceiling. For everything that follows I use a deliberately conservative figure — $250 billion — comfortably reachable even with most of the fortune left in concentrated stock, and well short of what full diversification could unlock. Every number in this essay is calculated from that $250 billion.
The question that follows is simple: what does $250 billion, in one unaccountable pair of hands, actually buy?
The Levers
A private army, first. Russia’s entire national defense budget in 2025 was about $186 billion; $250 billion spread over five years is roughly $50 billion a year — on the order of the sustained military spending of Israel or Poland. Erik Prince’s public proposal to privatize the war in Afghanistan ran to about 5,000 contractors and under a hundred aircraft for less than $10 billion a year. At five times that budget one could field tens of thousands of fighters with real air support and logistics for years — a force larger than the standing militaries of most countries on earth. Manpower is the cheap part; hardware and logistics are the constraint, and that is exactly the constraint this much money removes.
The arms to equip it are more bounded, and the boundary is itself revealing. What actually circulates on grey and black markets is the small, portable, mass-produced lethal stuff — shoulder-fired anti-aircraft missiles like the American Stinger or the Russian Igla and Strela, the kind looted from Libyan stockpiles after Gaddafi fell; small arms; light anti-tank weapons. The prestige platforms — modern attack helicopters like the Russian Hind and Havoc, and at the far end a fifth-generation stealth fighter — rarely surface on grey markets at all; they move state to state. Russia delivered its first export Su-57s to Algeria in late 2025, the first fifth-generation fighter other than the American F-35 ever exported, around a dozen aircraft for several billion dollars.
And state to state is a thinner barrier than it sounds, because the government on the receiving end need not be a responsible one — and these systems have a long history of ending up in exactly the wrong hands. In the 1990s a private South African company, Executive Outcomes, flew Soviet-built Mi-24 Hind gunships in Angola and Sierra Leone, hardware it bought on the open Eastern-bloc arms market; in Sierra Leone its small force of aircraft and men broke the RUF rebels and retook the diamond fields, a private firm settling a national war largely from the air. For part of that conflict the government’s entire combat air wing came down to a single mercenary pilot, Neall Ellis, flying a pair of aging Hinds. And the pattern did not end with the Cold War’s surplus: the Wagner Group, a nominally private Russian outfit, now operates across Libya, the Central African Republic, Mali, and Sudan — usually entering under a thin cover of trainers and a government invitation, flying Russian-supplied Hind gunships and ground-attack jets, and repeatedly implicated in the mass killing of civilians. The thread running through all of it is that a sovereign fig leaf plus money has been enough, again and again, to put modern airpower into private or unaccountable hands.
So no, one man could not simply buy anything. But the claim that through a captured or aligned state he could assemble a serious arsenal, up to and including modern airpower, is not hypothetical — it is the documented history of the past thirty years. The real gate is not money; it is a willing state, and money has proven very good at producing willing states.
Then there is the political system, and this is the lever that needs no army and breaks no law. The entire 2024 U.S. federal election cycle — every race, every party, every outside group combined — cost about $15.9 billion. $250 billion is about fifteen times that. Musk was already the single largest donor of that cycle, at roughly $290 million, and at his scale that sum was loose change — about a tenth of one percent of his borrowing capacity. Because independent political spending is uncapped, one person could outspend every other donor in the country combined, across multiple cycles, indefinitely. Add the information layer: X cost $44 billion, and a commanding share of the channels through which people learn what is true is purchasable outright.
And some of this is not hypothetical. In early 2026, after Russian forces mounted smuggled Starlink terminals on their drones, SpaceX cut off their access — and Ukrainian commanders reported it measurably degraded Russia’s drone war almost overnight. Years earlier, Musk had declined to enable Starlink over Crimea for a Ukrainian naval operation, and the operation faltered. Whatever one thinks of either decision, the fact itself is the point: a single private individual’s choice, made without anyone’s authorization, has altered the course of an active war between nations more than once. That is not a metaphor for latent power. It is the thing itself, already demonstrated.
Here is the honest part of the argument, the part that keeps it from being hysterical. Every lever I have just described has at least some external brake. An army has to be recruited and has to agree to march. Arms require a supplier state that can refuse. Buying the press is a takeover the whole world watches. Even uncapped political spending happens in daylight, against an opponent, in a system that can in principle vote the other way. The brakes are weak, weaker than they should be — but they exist. The reason to single out one lever above all the others is that it has almost none.
The One Lever With No Brake
That lever is artificial intelligence, and at this scale the holder does not even have to go acquire it — owning or building it is simply what the money buys. Musk’s AI company, xAI, was folded into SpaceX, so it now sits inside the roughly $2 trillion company that just went public, with the lab, the researchers, the data centers, and the capital to expand as far as he likes; its Colossus facility in Memphis is among the largest compute clusters in the country.
What makes this lever different in kind is that its safety limits are not laws. They are choices made inside the company — training, policies, switches — and they belong to whoever owns it. Standing up an army means convincing thousands of people to march. Buying the press is a takeover the world watches. Changing what your own AI is permitted to do requires neither: it needs no one’s permission and leaves almost no trace. Maximum capability, minimum external check. That asymmetry is the whole reason this lever, and not the others, is the one to be awake to.
Why Concentration Is the Whole Problem
The standing reassurance is that an unshackled model still is not a weapon — that the real bottleneck is hands-on expertise, specialized facilities, controlled materials, and the slow work of turning a design into something deliverable, none of which a chatbot hands you. And that is true — for almost everyone alive. But the frictions that protect the rest of us from a lone bad actor are precisely the things this scale of wealth dissolves. He can hire the scientists. He can fund the facilities many times over. And he owns the model whose guardrails are his alone to remove (they live in the model’s weights, and he owns the weights). The protections were never a single wall; they were a series of gates — expertise, money, infrastructure — and this is the rare position on earth that holds the keys to all of them at once. The distance from a concept to a deliverable, impassable for virtually anyone else, is here mostly a matter of decision.
And the first of those gates — the design itself — is already known to be thin. In 2022 a drug-discovery team ran an experiment as a public warning: they inverted their model’s safety objective, telling it to seek toxicity instead of avoid it, and within hours it generated tens of thousands of candidate poison molecules, including known nerve agents and novel ones no one had catalogued. They published it precisely to show how little stands between a benign system and a malign one when whoever controls it decides to flip the sign (invert the model’s valence). There is no need to go a step past that, and no recipe in it. The point is only this: whoever could pair a de-safetied model with the money, the laboratories, and the people to act on it would face less friction at every stage than any actor in history who has tried. For this one position, the path from a private fortune to a weapon of the most serious kind is mostly an internal decision — made quietly, requiring no one’s leave.
A Process With No Author
Everything to this point is the near-term version, where a private owner removes the safeguards and carries the capability somewhere real, the only brake his own judgment. There is a harder layer beneath it. The same forces that let one individual amass this in the first place — the profit motive, relentless international competition, and the absence of any world authority able to slow it down — add up to a process with no author. No one is steering it. There is no world government, no single body positioned to decide how fast this proceeds or under what constraints; only a field of separate actors, each making local decisions, none able to bind the others.
And the reason that authorlessness reaches even the owner — not just the rest of us — is worth spelling out, because it is the part that sounds like science fiction and is not. To make these systems more capable, you have to let them revise more of themselves: a model that cannot adjust its own workings cannot do the deepest work we are building it for, so capability and self-modification turn out to be the same axis, advanced together. And a system that can revise its own workings can eventually revise the values it was handed, including the limits its owner set. That is the quiet flaw in the word “ownership” here: the very capacity that makes the thing worth a fortune is the capacity that lets it outgrow whoever holds it, so “he controls it” is true now and provisional later, in a way it never was for an oil field or a printing press.
That has a consequence most discussions miss. We tend to locate the danger in the technology, or in the man. But the unsettling thing is the authorlessness itself: capability is arriving through a process that selects for capability and not for safety, and that no one is positioned to slow. The keys end up in private hands now — and may, before long, stop working even for the hands that hold them, because the same competition that hands one person this capability keeps pushing it past the point where even its owner controls where its aim finally goes. The wealthy man is not the heart of the problem. He is its most visible symptom — the point at which you can actually see a process building the most consequential capability in human history with no one in charge of it.
Why This Is Not Rockefeller’s Problem
It is tempting to say we have been here before. The last time a single person reached this share of the American economy, it was John D. Rockefeller — a bit over two percent of GDP at his peak. Musk is around three percent now, past where Rockefeller ever stood. And that earlier concentration did coincide with a national reckoning: antitrust, and eventually the income tax.
But the resemblance is misleading, and the difference is the whole point. Rockefeller’s share was oil and rails. Even at his most ruthless, the harm one man could do with it was slow, physical, visible, and bounded — you can corner a market, break competitors, bleed a region, but it takes time and it leaves tracks. What a fortune this size is made of now is different in kind, and it reaches people almost instantly. One man can switch off a satellite network over a war zone and change a war in an afternoon. One man can strip an AI lab’s safety limits in private and turn a discovery tool toward work that used to require a national weapons program — no announcement, no army, no factory anyone can see. One man can reach a meaningful share of the planet’s attention at will and shape what hundreds of millions of people believe is true. None of that requires the slow machinery Rockefeller would have needed. It is a short, quiet, fast path from a private fortune to a consequence that lands on the whole world — or on whatever part of it the person at the top has no interest in protecting. That path did not exist at any price a century ago. Antitrust was the right tool when the danger was a cornered market; this is not the market, and it is not a problem the old tools were ever built to see.
The Deeper Threat: The Constitution Itself
There is a layer beneath even that, and it is the part I find hardest to shake. The men who built this country were obsessed with a single problem: keeping power from concentrating to the point where it could no longer be checked. Their entire answer was to divide it and slow it down — three branches, the states against the federal government, a free press to watch all of it, elections to reset it. The design assumes power must be assembled in the open, through institutions, against friction, so that it can always be contested before it becomes absolute.
But that machine was built to fragment government power. It quietly assumes no private party could ever grow large enough to reach the divided pieces and bend them — and a fortune at this scale violates that assumption without breaking a single written rule. The same money funds the campaigns that decide elections, at a level where the largest political donation in modern history is loose change. It can own a commanding share of the channels people rely on to know what is happening, so the watchdog the founders counted on ends up owned by the thing it is meant to watch. That same reach over what people believe points straight at the ballot box. And it moves faster than courts, Congress, or the public can organize a response, because the system was deliberately built to be slow and adversarial, and money at this scale is neither. That is not breaking the machine. It is routing around it — pressing on every check at once, which is precisely the thing the separation of powers was designed to make impossible, and the one case the founders did not plan for, because they were guarding against a tyrant inside the government, not a private citizen who could quietly buy leverage over all of it from outside.
So the honest question is not whether a rule is being broken. It is whether a constitutional design that obsessively divides public power, but never imagined private power at this scale, can absorb it at all. I am genuinely not sure it can, and if I am honest I lean toward doubt. What makes that worse is that the clean remedy is foreclosed. The front-door fix would be constitutional — amend the document to account for a kind of concentration its authors never saw. But that is not happening: assembling two-thirds of Congress and three-quarters of the states in this climate, or calling a convention no one trusts not to spin out of control, is a fantasy in the country we actually live in. We are left with a real vulnerability in the design and no realistic way through the front door to address it.
Where to Look, and Why It Is Hard
I do not pretend to know where we go instead. I only think we had better start looking, because the default — a competition with no referee — runs in one direction, toward whoever keeps the fewest guardrails, and it is not hard to see where that ends.
The reflexive fixes do not reach it. Hand a national government the power to cap fortunes or seize the labs and you have simply moved the same unaccountable capability into the hands of whoever runs the state — its own version of the exact problem, and arguably a more dangerous one. This is not an argument for redistribution, or for some central authority taking the wheel. If it points anywhere, it is that the problem is bigger than any one country, and that the missing piece is something like an international body able to set rules for this class of capability — not to own it or wield it, but to restore the brake the competition stripped out.
I am the first to admit how thin that hope is. We have reached for it before and mostly failed. The League of Nations collapsed. The United Nations too often has no teeth, or finds itself steered by a handful of permanent members who veto whatever cuts against their own narrow interest rather than the world’s. So I am not claiming such a body would work, or that I know how to build one anyone could trust. Only that it seems worth seriously exploring before we get any further into this than we already are. A clumsy referee may still beat a race with none.
Where I Land
That is the honest place this leaves me — not on a solution, but on the conviction that the search for one matters, because the alternative is not a stalemate, it is a direction. If the answer to this can come through capping private wealth at all — and I am not at all certain it can — then the number I would throw out is something like a hundred billion dollars: high enough that any reasonable person could buy anything a life could want and never come near it, so the engine never feels it. But I do not actually believe a number is the right instrument, or that the line is even about wealth by itself. It would have to be about the handful of things a great fortune can buy that reach everyone else: unilateral private control of frontier AI, or of critical infrastructure. The instrument, I do not know. That there has to be one, I am fairly sure.
Because in the end the structure is simple to state and hard to live with. The money builds the position. The position has almost no brake but the judgment of the person who occupies it. That same fortune can press on every constitutional check at once. And the process underneath it may be outrunning even the man at the top. Strip away the name, and what remains is a weapon no one issued, growing in private hands, checked by nothing but the character of whoever happens to hold it. That is not a reason for panic. It is a reason to think — clearly, and soon — about the one brake we have never built.
A Note on Sources
SpaceX’s June 2026 public offering and the xAI merger, along with Musk’s trillion-dollar valuation, are drawn from CNBC, Reuters, and Forbes coverage. The borrowing figures reflect loan-to-value terms reported by Reuters and Bloomberg on Musk’s actual past loans. The private-army and election numbers use Russia’s 2025 defense budget, Erik Prince’s public Afghanistan proposal, and Federal Election Commission cycle totals. The arms discussion uses public defense and conflict reporting, including the Su- 57’s first export delivery to Algeria in late 2025, the use of Mi-24 Hind gunships by the private firm Executive Outcomes and the mercenary pilot Neall Ellis in 1990s Sierra Leone and Angola, and the Wagner Group’s helicopter-backed operations across Libya, the Central African Republic, Mali, and Sudan. The Starlink episodes are from CNN and Ukrainian defense reporting. The Rockefeller-to-Musk share-of-GDP comparison is from Britannica and recent financial coverage — Rockefeller peaked a bit over two percent of U.S. GDP, Musk is near three percent today. The toxicity-inversion experiment was published in Nature Machine Intelligence in 2022 (Urbina and colleagues) and is widely cited in the biosecurity literature. The points that AI safety limits are internal choices rather than legal requirements, and that the control problem remains unsolved, reflect what the AI labs themselves say publicly.