This is Part 2 of a two-part look at threats posed by frontier AI development, capitalist accumulation, and the logic that keeps us awake at night.

I recently watched a Breaking Points interview with Nate Soares. Asked about the possible motives behind multiple AI-tech CEOs issuing public warnings about a potential AI-caused cataclysm, he suggested a structural pressure at work. Employee researchers confronting the difficulty of making AIs follow instructions and not commit cybercrimes are spooked. AI corporate leaders may be trying to say enough to head off revolt and more resignations by employees who don’t want advancing a catastrophe to be “in their personal hands,” while saying little enough that regulators do not step in.

Watching Soares talk brought me back to an argument I made recently about Jacob Coxon’s resignation from Anthropic. My claim then was twofold. First, that the race toward self-improving superintelligence despite acknowledged existential risk is not some bizarre deviation from capitalism’s normal operation, but an unusually stark expression of its immanent logic: compete, accumulate, expand capability, keep moving because someone else will if you do not. Second, that the very capacities Coxon asks us to fear in advanced AI—the ability to “hack anything, revolutionize any field… acquire real power and resources”—are unsettling in part because they so closely resemble the logic of capital itself.

But that argument, and Soares’s intuition about AI employee anxieties impacting CEO public posturing, opened the door on another question. If the force of capitalism lies partly in its ability to make historically produced imperatives appear inevitable, where does resistance enter? What allows people not merely to recognize the logic they are caught inside, but to interrupt it?

I know from experience, on a vastly smaller scale of consequence, something about the dilemma Soares describes. I once worked in an institution whose frontline staff increasingly believed that a change in leadership was carrying the organization away from the mission for which it existed and toward practices some of us regarded as ethically indefensible. Yet leaving was not an uncomplicated act of integrity. Those of us most committed to the mission were also, inconveniently, among the people still making the institution behave as though it had one. If we left, we had every reason to believe we would be replaced by people either less troubled by the new direction or more willing to accommodate it. “If not me, who?” can be a wonderfully serviceable rationalization for staying inside a compromised institution. The trouble is that sometimes it is also true.

This is one of those tricky places where we need to be careful about the shadow allure of structural analysis. Such analysis shows us how choices made by actors are reasonable given the structural cross-pressures at play. The researcher who remains inside a frontier AI lab because somebody responsible ought to be there has reasons. The executive who insists that slowing down would merely surrender the field to a less responsible competitor has reasons. The policymaker frightened that regulation will hand technological advantage to China has reasons. Capitalism is exceptionally good at generating reasons why the next step must be taken.

Structurally, there are always reasons.

But we are under no obligation to evaluate those reasons only in their own terms. Structures explain why refusal is costly; they do not make refusal impossible. Somewhere between necessity and inevitability sits the troublesome human capacity to say: “Yes, I understand perfectly well why I am supposed to keep doing this. I am not going to.“

Melville gave American literature a dazzlingly pure version of this workplace refusal in Bartleby the Scrivener’s “I would prefer not to”—in, of all places, A Story of Wall Street. But Bartleby’s refusal remains solitary and finally self-consuming; the harder political question is how “I would prefer not to” becomes “we will not.”

And here is where courage enters—not as some heroic personality trait floating free of circumstance, but as a genuine variable in what otherwise looks like an inexorable process. Coxon’s resignation matters first at this elementary level: somebody confronted the reasons for remaining, judged them insufficient, and accepted the costs of refusal. But he did something more consequential than quietly clear out his desk. He explained himself in public. In doing so, an individual act of conscience became a high-visibility event: an insider with unusual knowledge of the work announced that the alarming interpretation of what was happening was serious enough that he would no longer place his own labor behind it. A private moral decision became public witness—and that begins to change the field in which everybody else makes theirs.

We also have a ready illustration of what can happen when private refusal becomes collective refusal. In 2019, the editorial staff of online sports publication, Deadspin, found itself in a conflict with new ownership over the purpose of the publication itself. Management issued a now infamous instruction to “stick to sports,” fired its acting editor after the staff resisted, and over the next several days the editorial staff quit en masse.

The important thing is not that a collection of unusually principled journalists individually decided to resign. It is that they were able to transform a series of personal moral choices into a collective act. “I cannot keep doing this” became “we will not keep doing this.” That transformation matters because the act of organization changes the menu of reasons.

As we have seen, the individual worker confronting a compromised institution may have powerful reasons to stay. Those reasons were surely real for people inside Deadspin, just as they assuredly are inside AI labs now. But collective action creates reasons that did not previously exist. If enough of us refuse together, perhaps management cannot simply replace us one by one. Perhaps the institution’s dependence on our labor becomes visible. Perhaps the thing we care about is not identical with the corporate shell that currently owns it. Structures explain why courage is costly. Organizing changes what courage can accomplish.

The Deadspin case makes that last possibility unusually vivid because the story did not end simply with refusal. Eighteen former staffers went on to create a new publication, Defector, as a worker-owned company, explicitly in response to their experience under private-equity ownership. They did not return individually to some new version of the same institutional arrangement. They built another one.

The work they cared about turned out not to be inseparable from the ownership structure claiming possession of it. The journalism, accumulated skill, institutional memory, relationships among coworkers, voice, and relationship with readers could walk out the door.

And what happened afterward matters too, though not because it supplies a tidy capitalist morality tale in which the virtuous workers went off and beat the bad owners at their own game. (And it is decidedly not my aim here to assess alternatives to capitalist logic in capitalism’s own terms of success.) The more important fact is that Defector proved materially viable. Five years in, it remains worker-owned, employs more people than it did at launch, has tens of thousands of paying subscribers, and has maintained a remarkably stable operating business in the ruinous economics of digital media. Its own annual reports are notably unsentimental about the difficulty of doing so. Worker ownership is no magic funding machine; what Defector has demonstrated is something more interesting: a social practice can reproduce itself materially without subordinating its reason for being to an external imperative of private accumulation.

The fate of Deadspin itself gives us a split-screen view of almost indecent clarity. The brand survived. The institution, in the sense that mattered to the people who had made it what it was, largely did not. The name eventually changed hands again; today the site is filled heavily with betting picks, odds, affiliate offers and links into the sports-gambling economy. Capital retained the property and found new ways to extract value from the residue. The workers, meanwhile, had carried away much of the productive social institution and rebuilt it somewhere else.

Ownership had legal possession of Deadspin, but legal possession of the asset was not the same thing as possession of the social relations that made the asset valuable. Once those relations became inconvenient, they could be discarded while the name, traffic, recognition and accumulated reputation remained available for further monetization. The corporate shell survived; much of the institution worth preserving walked out the door.

And this is where resistance begins to become something more than conscience. Refusal can reveal that an imperative is not inevitable. Collective refusal can impose costs on the institution enforcing it. Organizing can change the available choices. And institution-building can go further still: it can create a social form in which different purposes govern what people do. The revolutionary moment, in miniature, is not simply when people stop obeying the old logic. It is when they discover that what appeared to require that logic can be made materially viable under another one.

The revolutionary moment begins when we recognize that what had appeared as something like weather conditions reappears as nakedly human arrangements of power

Now let’s carry this discovery of contingency back into the AI laboratories. The moral problem confronting frightened researchers does not have to remain a private accounting exercise in which each person asks whether staying or leaving will do marginally less harm. Coxon’s public resignation matters because it helps make what might otherwise remain scattered private apprehensions mutually visible. The moment people begin to recognize that they are not separately experiencing the same dilemma but confronting a common condition, what had appeared as something like weather conditions reappears as nakedly human arrangements of power. And a question of conscience can become a question of collective power.

But AI is only a particularly vivid instance of a larger crisis. One of capitalism’s great achievements has been to make its own requirements feel less like the demands of a particular social order than like natural facts about reality. We may regret the consequences of the choices we are bound to make, but what choice do we have? There are always reasons. Arrangements produced by human beings return to confront human beings as necessities to which we must adapt. The market becomes weather. Competition becomes gravity. Accumulation becomes the price of existence. And political argument is reduced to a discussion of how comfortably we might arrange ourselves inside forces we have somehow agreed to imagine are beyond political reach. They are not.

Capitalism’s imperative is to convert ever more of human social life into fields of extraction for private accumulation. It will never peacefully coexist with a way of life organized for the development of human social well-being. It cannot. This is why the crisis cannot be addressed by simply finding more enlightened owners, more conscientious executives, more socially responsible investors, nor even by regulatory regimes subject to capture by the familiar corporate imperatives. The answer cannot simply be gaining better control by or over the same class occupying the commanding positions of the same machinery. It has to include changing the machinery that keeps manufacturing the same structures, the same arrangements of power, and the same naturalization of the reasons why things must remain so arranged.

This is what makes the modest Defector example more than an inspiring workplace anecdote. Those workers did not persuade private-equity ownership to become nobler. They did not win a philosophical debate about journalism. They discovered that the activity they valued could be separated from the ownership relation claiming authority over it, carried elsewhere, reorganized around different purposes, and sustained. For one tiny piece of the social world, the supposed necessity dissolved. That is a far more radical fact than whether they subsequently beat some competitor in clicks, valuation, or return on investment. It says that the social activity and the capitalist arrangement enclosing it were not the same thing.

Once that distinction becomes visible, a great deal else becomes visible. Hospitals are not the same thing as healthcare markets. Universities are not the same thing as tuition streams and endowments. Housing is not the same thing as real-estate appreciation. Knowledge is not merely proprietary information waiting to be monetized. And the accumulated language, art, science, memory and imagination of humanity are not, by some law of nature, taken up as raw material awaiting conversion into privately owned machine intelligence. These are arrangements. They have histories. They are maintained by institutions of our making, laws, concentrations of power and, not least, by our repeated and habitual participation in them. Which means they can be contested.

And here I think the word revolutionary deserves to be recovered from both adolescent romance and respectable terror. Revolution need not mean that one morning everybody wakes up, storms the Winter Palace and discovers that history has changed sides just in time for happy hour. It can mean something at once more ordinary and more demanding: human beings recovering the power to decide what the institutions of their common life are for, who owns them, whom they serve, what purposes they answer to. We can recover the power to decide what we will allow to be bought and sold. What we will insist remain common. What forms of power we will no longer consent to treat as private prerogative.

The revolutionary possibility appears whenever people stop accepting the available choices as the boundaries of political imagination and begin altering the arrangements that produce those choices in the first place. It appears as the recognition that there is always an alternative.

It may be—will be—difficult. It will be costly. It may require surrendering advantages, fortunes, careers, institutions, habits and entire ways of understanding what economic life is for. It will require organizing power strong enough to overcome people who benefit enormously from the arrangement as it stands. Nothing about calling an arrangement contingent makes changing it easy.

But difficult is not inevitable. Costly is not impossible. Powerful is not natural law.

The prospect of runaway artificial intelligence frightens us because we imagine a system pursuing its objective with enormous capability, absorbing resources, circumventing barriers and relentlessly extending the logic of its goal into every domain it can reach, even when the consequences become destructive of the human world around it. But we have been living with that scary nightmare all along. The society building the machine already contains a system that behaves this way. We have been inside it, living in the house of capitalism.

But as we have seen, even from inside the house we can organize our refusal. We can find one another. We can withdraw our labor, our legitimacy, our obedience and our consent. We can defend what capital would consume. We can refuse the conversion of every human need into a market, every human capacity into an asset, every commons into an enclosure, every relationship into a transaction, every new technology into another engine for concentrating private power.

And once organized, as we have seen, we can build the alternative. We can build institutions whose purpose is not accumulation but the enlargement of human capacity to make a richer, safer, healthier human home in the world. And we can build power sufficient to protect them.

Capital insists that its imperatives are inevitable. They are not. They are enforced.

And what is enforced can be resisted.