Extinction-level capitalism a citizen’s thoughts
on AI risk

AI is inher­ently polit­ical tech­nology. If AI works as intended, it will grad­u­ally corrode our liberal democ­racy, risking an irre­versible shift into another polit­ical and economic config­u­ra­tion. Among AI risks, this one deserves more consid­er­a­tion, because it requires no addi­tional condi­tions like malign actors or AI malfunc­tion. AI only needs to amplify existing trends, espe­cially around concen­tra­tion of capital. This damage will occur even assuming that in the near term, AI will broadly improve mate­rial well-being.

About MB

I’m a self-employed author, designer, programmer, and lawyer. In 2022, I learned that my own works were in the training datasets of gener­a­tive-AI compa­nies. In response, I invented the first set of lawsuits chal­lenging the legality of these prac­tices. I’m currently co-counsel for plain­tiffs in a number of AI cases. Though I discuss certain legal issues below, I am not your lawyer, and nothing here is held out as legal advice. These are my personal views as a citizen and economic actor; I speak only for myself. This piece is typeset in Equity, Advo­cate, and Trip­li­cate, fonts I designed. They can be licensed for your own polemics and pamphlets.

Emergent effects

Two billion years ago, the rock layers comprising what is now called the Colorado Plateau began to form: first igneous and meta­mor­phic rocks, followed by many layers of sedi­men­tary rocks. About fifty million years ago, through tectonic action, this plateau gained thou­sands of feet of eleva­tion. About five million years ago, a river began to flow. The river carried silt and debris, scraping out the begin­nings of a canyon. The river deep­ened the canyon, exposing its walls to weather and erosional forces that widened the canyon further. Today the waterway is the Colorado River. The geolog­ical forma­tion is the Grand Canyon.

The forma­tion of the Grand Canyon required zero human agency. Zero tech­nology. Zero coor­di­na­tion among the river, the land, and gravity. In that sense the Grand Canyon is an emer­gent effect: a complex, unfore­see­able output arising from simpler inputs.

But we would never wonder whether the river is sentient. Or whether the river cares about the dirt that it carries out of the canyon. The water is just doing what water does: flowing down­hill. The dirt just happens to be in the way.

Inherently political technology

Langdon Winner is a polit­ical theo­rist. Winner wrote the excel­lent and influ­en­tial essay “Do Arti­facts Have Poli­tics?” (1980). Winner sought to debunk the tradi­tional framing that “tech­nolo­gies are … neutral tools that can be used well or poorly, for good, evil, or some­thing in between.” Instead, Winner proposes two ways that a tech­nology can affect its polit­ical envi­ron­ment:

1. The tech­nology is designed to have certain polit­ical effects. For example, the Great Fire­wall of China, a bundle of tech­no­log­ical measures that limit Chinese citi­zens’ access to foreign infor­ma­tion sources. Antipo­dally, the Tor Project intends to maxi­mize user anonymity and thwart govern­ment intru­sion.

2. The tech­nology is inher­ently polit­ical. This is Winner’s key analytic fulcrum. Winner describes two versions of inher­ently polit­ical tech­nology. The first is where the tech­nology “actu­ally requires … a partic­ular set of social condi­tions as [its] oper­ating envi­ron­ment.” For instance, nuclear weapons: the only respon­sible way to possess such dangerous tech­nology is to place it within “a central­ized, rigidly hier­ar­chical chain of command … the [atom] bomb must be author­i­tarian; there is no other way.” The second version is where the tech­nology is “strongly compat­ible” with a certain polit­ical arrange­ment (even if not strictly required) and thus tends to bring that arrange­ment to fruition.

As an example, Winner considers the mechan­ical tomato harvester. Devel­oped at UC Davis in the 1950s, the machine was tremen­dously produc­tive. But it was also expen­sive. Only well-capi­tal­ized tomato growers could afford it. The rest couldn’t compete. According to Winner, the number of Cali­fornia tomato growers dropped from ~4000 in the early 1960s to ~600 in 1973, costing ~32,000 jobs and the compounding nega­tive effects on those commu­ni­ties. Winner summa­rizes:

What we see here … is an ongoing social process in which scien­tific knowl­edge, tech­no­log­ical inven­tion, and corpo­rate profit rein­force each other in deeply entrenched patterns that bear the unmis­tak­able stamp of polit­ical and economic power … oppo­nents of inno­va­tions like the tomato harvester are made to seem “antitech­nology” or “antiprogress”. For the harvester is not merely the symbol of a social order that rewards some while punishing others; it is in a true sense an embod­i­ment of that order.

Not merely the symbol—the embod­i­ment. A facially neutral tech­no­log­ical inven­tion—here, a tomato harvester—can induce polit­ical effects. Those effects don’t arise from flaws in the tech­nology. To the contrary—they arise from its effi­cacy.

How are the polit­ical effects deter­mined? Winner iden­ti­fies two key early deci­sions. The first is the binary ques­tion of whether to pursue the tech­nology at all. The second are choices about “the design or arrange­ment” of the tech­nology. Winner cautions: “[t]o see the matter solely in terms of cost-cutting, effi­ciency, or the modern­iza­tion of equip­ment is to miss a deci­sive element”. That is, the polit­ical effects can possibly be coun­tered, but first they must be acknowl­edged.

Of course, the best oppor­tu­nity to choose wisely is before the tech­nology is widely intro­duced, as capital and social invest­ment tends to entrench it:

Because choices tend to become strongly fixed in mate­rial equip­ment, economic invest­ment, and social habit, the orig­inal flex­i­bility vanishes for all prac­tical purposes once the initial commit­ments are made. In that sense tech­no­log­ical inno­va­tions are similar to legisla­tive acts or polit­ical found­ings that estab­lish a frame­work for public order that will endure over many gener­a­tions. … The issues that divide or unite people in society are settled not only in the insti­tu­tions and prac­tices of poli­tics proper, but also, and less obvi­ously, in tangible arrange­ments of steel and concrete, wires and tran­sis­tors, nuts and bolts.

Tech­no­log­ical choices bear directly on the “public order” at large. When we don’t take those choices seri­ously—or we’re persuaded to ignore them by those insisting that tech­nology is just a neutral tool—we risk polit­ical conse­quences.

Winner warns of compla­cency. Once the tech­nology arrives and becomes entrenched, the conver­sa­tion gets reframed as one of tech­no­log­ical inevitabilism vs. anachro­nism, and dissent is discour­aged: “the kinds of reasoning that justify the adap­ta­tion of social life to tech­nical require­ments pop up as spon­ta­neously as flowers in the spring … After a certain point, those who cannot accept the hard require­ments and imper­a­tives will be dismissed as dreamers and fools.”

Liberal democracy

The balance of power of democ­racy is premised on the average person having leverage through creating economic value. If that’s not present, I think things become kind of scary.
—a certain AI CEO

Liberal democ­racy is the polit­ical scien­tist’s term for the type of govern­ment preva­lent among capi­talist economies since the Amer­ican and French Revo­lu­tions. The intel­lec­tual foun­da­tion of liberal democ­racy arose during the Enlight­en­ment, espe­cially through the work of John Locke. Liberal democ­racy empha­sizes limited govern­ment, indi­vidual rights, and sepa­ra­tion of powers—in short, majority rule with excep­tions and guardrails. (The term liberal democ­racy doesn’t connote liberals or Democ­rats in the specific US polit­ical sense. But polit­ical parties of differing ideolo­gies are a tradi­tional feature of liberal democ­ra­cies.) Today, most liberal democ­ra­cies are in Europe, the Amer­icas, and the Pacific Rim.

Liberal democ­racy is not a fixed set of immutable char­ac­ter­is­tics, but a bundle of graded values. All liberal democ­ra­cies empha­size certain ones over others. In the aggre­gate, some of these nations evolve toward stronger liber­alism; others evolve away. These degraded cases have some­times been called illib­eral democ­racy: the observ­able formal­i­ties of liberal democ­racy may still be observed—e.g. multi­party elec­tions, sepa­ra­tion of powers—but the lived reality is single-party rule and declining indi­vidual rights.

That’s not to say that liberal democ­racy produces excel­lent outcomes for all citi­zens, all the time. It doesn’t. At any moment, certain citi­zens are dissat­is­fied—say, because they belong to a group whose rights are inad­e­quately protected or econom­i­cally margin­al­ized. Liberal democ­racy offers a process, not a result: grass­roots demo­c­ratic partic­i­pa­tion can coalesce into policy change. But within the arena of competing polit­ical inter­ests, winners and losers neces­sarily follow. Navi­gating these differ­ences within a stable, accom­moda­tive frame­work is prefer­able to a rigid one that buckles under these stresses—say, through polit­ical revo­lu­tion, which tends to be messy and unpre­dictable.

Capi­talism has tradi­tion­ally been consid­ered a neces­sary but not suffi­cient condi­tion for liberal democ­racy. Why? A regu­lated market economy encour­ages citizen partic­i­pa­tion through prop­erty owner­ship and trans­ac­tion. A prin­cipal func­tion of govern­ment is to define the economic condi­tions of the state. Economic partic­i­pa­tion mutu­ally rein­forces demo­c­ratic partic­i­pa­tion. Prop­erty owners will vote for those who protect their inter­ests. The rise of indus­trial capi­talism in the 19th century, and the wealth-redis­tri­b­u­tion mech­a­nisms that followed in the 20th, led to economic empow­er­ment for more citi­zens, and ulti­mately broader polit­ical empow­er­ment. The converse also holds: economies premised on state or oligarchic control of some narrow class of assets haven’t tended to evolve toward liberal democ­racy.

In prac­tice, certain people in a capi­talist liberal democ­racy tend to get increas­ingly rich. Absent coun­ter­mea­sures, the wealthy gain control of the polit­ical appa­ratus, thwarting liberal-demo­c­ratic norms. This tension between capital and poli­tics is a long-consid­ered topic. A key early work was, of course, Karl Marx’s Capital (about which more later). In the current era, Mancur Olson’s book The Rise and Decline of Nations set out how small groups with a shared interest (which could include capital concen­tra­tion) can effec­tively under­mine stable soci­eties. More recently, econ­o­mists Robert Reich (“How Capi­talism is Killing Democ­racy”), James Galbraith (The Predator State), and Yanis Varo­ufakis (Tech­nofeu­dalism: What Killed Capi­talism) are among those who have studied the esca­lating polit­ical conse­quences of rising wealth inequality. The synthesis might be: as more wealth becomes concen­trated in the hands of fewer citi­zens, liberal democ­racy weakens, because whichever citi­zens are losing economic rele­vance will also lose polit­ical rele­vance. A nation sending many of its citi­zens toward economic irrel­e­vance risks becoming polit­i­cally illib­eral.

The Skynet fallacy

AI discourse often invokes sci-fi narra­tives. I’ve called this the Skynet fallacy, after the Termi­nator antag­o­nist, the most cited. But any sci-fi will do. For instance, one AI CEO warned of AI “going Termi­nator”; Stephen Hawking and other scien­tists warned of AI “devel­oping weapons we cannot even under­stand”; a second AI CEO said we “don’t need much imag­i­na­tion [about AI risk] because we grew up with that in the media”; a third AI CEO invoked the sci-fi movies Contact and 2001: A Space Odyssey in a piece about AI risk; a US congressman summa­rized AI risk as “evil robots rising up to take over the world”; a well-known jour­nalist advo­cated for more Termi­nator analo­gies; a promi­nent AI-risk pundit said he’s “annoyed” with Termi­nator analo­gies yet has suggested that AI will erad­i­cate humanity using hordes of toxic nanobots.

Artis­ti­cally, sci-fi movies exter­nalize the awe and unease of tech­no­log­ical confronta­tion. It’s easy to see why these metaphors have become part of AI-risk discourse. And yet. As AI puts down roots in our economy, the sci-fi framing hides more than it reveals. Sci-fi plots are opti­mized for cine­matic impact. So as a metaphor for AI risk, they can lead to faulty intu­itions. Among real­istic AI risks, we can expect that most will be boring, slow, and depend on minimal extra tech­nology. Whether AI will cause literal human extinc­tion is esoteric—a light­ning strike. But AI could easily induce future economic and polit­ical condi­tions that most Amer­i­cans today would consider intol­er­able—a cancer that extin­guishes a certain way of life. Nobody’s going to make a movie about boring AI risks. But they comprise the majority of worri­some AI outcomes.

In 2003, philoso­pher Nick Bostrom proposed his now well-known parable of the paper­clip maxi­mizer. Bostrom imag­ines an advanced AI that is asked to make paper­clips. Taking its mission seri­ously, the AI erad­i­cates humanity as it consumes all Earth resources to make more paper­clips. Bostrom was illus­trating the AI control problem: ensuring an AI acts consis­tently with human prior­i­ties is diffi­cult, even for osten­sibly simple goals. Because when we say “AI, make paper­clips”, the implied coda is “… without killing everyone.” But an AI can’t know this ex nihilo. Bostrom softens the sci-fi flavor by choosing paper­clips and not, say, laser-wielding robots. Bostrom’s choice of an economic mech­a­nism of resource conver­sion is apt. Even a mundane objec­tive can produce outsize risk. We could further observe that on the paper­clip-maxi­mizing path, human life would become dystopic long before literal extinc­tion. As resources are depleted, humans would become tenants in a neofeu­dalist paper­clip empire. (Paper­clip Crisis: The Saga Begins—opening soon.)

Computer scien­tist Stuart Russell also explored the control problem in his book Human Compat­ible. Russell calls one variant the gorilla problem: that “ances­tors of the modern gorilla created … the genetic lineage leading to modern humans. How do the gorillas feel about this? … the consensus opinion would be very nega­tive indeed.” Despite shared lineage, gorillas and humans have incom­men­su­rable values. Nothing humans can do—short of disap­pearing—would restore gorillas to their golden age. Russell’s framing is a believ­able analogy for the future rela­tion­ship of humans and AI. Sure, gorillas didn’t delib­er­ately invent humans. But I take Russell to mean that the emer­gent char­ac­ter­is­tics of these rela­tion­ships are more conse­quen­tial than the intended ones. (In that sense, Russell echoes Langdon Winner.) The fact that we’re inventing AI doesn’t mean we will predict or control its gravest effects. Any more than gorillas could predict or control human domi­nance of their ecosystem. The gorillas did their thing. We did ours.

AI will do its thing too. It will take time to figure out what that is, exactly.

How the West was won

I recently read Cadillac Desert by Marc Reisner, about the devel­op­ment of water resources in the western US between 1910 and 1980, espe­cially the dam-building campaign of the federal Bureau of Recla­ma­tion. Reisner’s book weaves several story­lines:

Engi­neering and envi­ron­mental impacts. Dams concen­trate water. But they cause envi­ron­mental conse­quences else­where. Further­more, Recla­ma­tion’s 20th-century projects were often based on opti­mistic projec­tions of mete­o­ro­log­ical water supply. Today, long-term drought condi­tions chal­lenge those projec­tions. No amount of money or hydro­logic engi­neering can change that.

National US poli­tics. In the early 20th century, newer Western states sought polit­ical clout in the federal govern­ment, which had been domi­nated by Eastern states. Polit­ical clout followed economic growth, and to achieve growth, the Western states depended on one crit­ical but scarce input: water. In the West, the federal Bureau of Recla­ma­tion was tasked with increasing water supplies. A polit­ical symbiosis emerged: congressmen from Western states voted to fund Recla­ma­tion water projects; in turn, Recla­ma­tion looked out for their constituents. Over decades, Reisner depicts this rela­tion­ship as metas­ta­sizing from prac­ti­cality to corrup­tion, in the sense of Recla­ma­tion becoming beholden to a narrow polit­ical lane. In that sense, Recla­ma­tion’s dams arguably qualify as inher­ently polit­ical tech­nology.

State economies. Recla­ma­tion’s highly subsi­dized water boot­strapped Western economies, espe­cially agri­cul­ture. For a while, it worked as promised: Farmers got irri­ga­tion. Cities got water. Western states grew and pros­pered. But the projects worked so well that Western states wanted more. These states never weaned them­selves from subsi­dized federal water, setting them on a path toward perma­nent depen­dence.

The parallel between water and AI is inexact. Water is a biolog­ical neces­sity; AI is not. Recla­ma­tion’s projects worked (up to a point); AI may or may not. This is part of why AI propo­nents have sought to raise the stakes. So far, AI has been grue­somely expen­sive and deliv­ered middling results. Never­the­less it’s routinely depicted as a geopo­lit­ical fulcrum, a proxy for contin­uing US excep­tion­alism. If Amer­i­cans don’t adopt AI whole­heart­edly, we will be losers. Do you want to be a loser?

Labor replacement

Q: What is Big AI primarily selling? A: Labor replace­ment, with mass unem­ploy­ment as a likely conse­quence. Some disagree or call it doomerish. Why? It’s exactly what AI grandees have been telling us. A certain AI CEO wrote that AI “will be hugely desta­bi­lizing for hundreds of millions” and that AI tools “are funda­men­tally labor replacing”. A certain AI company released a research paper about “the labor market impact poten­tial of large language models”. That AI CEO said “jobs are defi­nitely going to go away, full stop”. A second AI CEO said that in the near future, “20% of people don’t have jobs.” A third AI CEO predicted that farther out, “prob­ably none of us will have a job.” An AI-adja­cent CEO said that AI “will destroy human­i­ties jobs”. The ball is not hidden.

Capital markets are already pricing in these expec­ta­tions. Regard­less of whether Big AI even­tu­ally delivers mass labor replace­ment, today these compa­nies seek to concen­trate capital as if they will. According to the Wash­ington Post, AI capital expen­di­ture in 2026 is esti­mated to be $700 billion, a “spending spree [that] has few prece­dents”. Based on a recent survey of US workers, the Global Part­ner­ship on Arti­fi­cial Intel­li­gence said that “the policy window to shape how AI trans­forms work is prob­ably closing faster than most govern­ments realize.”

Extra­or­di­nary invest­ment demands extra­or­di­nary returns. In early 2026, after a certain billion­aire tech CEO laid off 40% of his employees and attrib­uted the deci­sion to a new “core thesis” of AI, the company’s stock rose nearly 25%. He won’t be the last. Whether these layoffs are based on actual AI bene­fits or merely “antic­i­pa­tory” is neither here nor there. Employers have strong incen­tives to reduce head­count and increase AI spending before competi­tors do. We will increas­ingly see both kinds of layoffs. Soft­ware program­mers are one set of conse­quen­tial, highly paid writers likely to be replaced with AI. Else­where I’ve predicted that legal prac­tice will also be seri­ously impaired. Why? Like program­mers, lawyers write about conse­quen­tial things and charge a lot. (We can expect that AI compa­nies them­selves, fond of dogfooding, will perfor­ma­tively set an example through their own AI-driven layoffs.)

Two common objec­tions to this conclu­sion:

1. “AI won’t replace workers, it will enhance them.” That may turn out to be true. But for now, it’s not what AI CEOs are pitching. And it’s not how the market is valuing Big AI. AI customers will natu­rally pay less for a worker-assis­tive tech­nology than tech­nology that reduces head­count, which is one of the biggest expenses at any employer. The labor-replace­ment story is every­thing to every­body. If AI is just a neat way of, say, finding photos of your dog on your phone, none of the AI bets pay out.

2. “AI will replace certain jobs, but also create new ones.” Fair, but “create new ones” is skimpy on details. How do the jobs created compare to the ones lost? Do they pay more? Or less, as the supply of workers seeking work increases? Do the new jobs require different skills? Who bears the cost of the career switch? None of this is auto­matic. As econ­o­mist Carl Benedikt Frey said: “Most econ­o­mists will acknowl­edge that tech­no­log­ical progress can cause some adjust­ment prob­lems in the short run. What is rarely noted is that the short run can be a life­time.”

After going hard on this narra­tive for several years, AI CEOs are now soft­ening their tone, primarily in response to public back­lash. For instance, the AI CEO who once said “jobs are defi­nitely going to go away” now wants us to know that watching AI answer his emails felt “dehu­man­izing” and that he now values “the human part of the roles”. That’s what it finally took—using your own product? Ed Zitron has char­ac­ter­ized the AI market so far as a blink­ered conver­sa­tion between AI exec­u­tives and trad exec­u­tives, all of whom “have little idea what work looks like”. This is more a feature than a bug. To replace labor, AI doesn’t need to actu­ally deliver compa­rable outputs. It only needs to fulfill a certain exec­u­tive fantasy of “what work looks like”. That’s much easier than, say, achieving AGI (however defined). On that view, AI may turn out to be a Veblen good—an item that buyers value primarily for signaling their ability to afford it.

One more small matter—funding of the US govern­ment is premised on employ­ment. In 2023, the US got 48% of its revenue from income taxes and 36% from other payroll taxes. If AI causes 30% unem­ploy­ment, revenues will drop and expenses will jump. Of course, higher unem­ploy­ment also leads to lower aggre­gate consumer spending. These would be fiscal condi­tions unknown since the Great Depres­sion, and utterly foreign to Amer­i­cans living today.

The goodies economy

Big AI’s message to poten­tial corpo­rate customers of AI has centered on labor replace­ment. Correctly fore­seeing that the workers them­selves might take umbrage, Big AI has pitched them a sepa­rate story: that workers will enjoy higher stan­dards of mate­rial well-being under AI.

One billion­aire AI CEO predicted “universal high income” and that “AI and robots will provide any goods and services that you want.” Another billion­aire AI CEO predicted “universal extreme wealth for every­body”. A billion­aire AI venture capi­talist “believe[s] in making everyone rich, every­thing cheap, and every­thing abun­dant” and that “the ulti­mate result will be that all phys­ical goods become as cheap as pencils.” A tech billion­aire predicted “more goods and services with less labor” and “always-avail­able, high-quality medical advice.” Another billion­aire AI venture capi­talist predicted “the abun­dance of goods and services … will be very, very large. Prices will be very, very, low.” Hope­fully yachts will come down too.

But wait—will they? In the post-AI economy, will we all have yachts? Of course not. As usual, we should be skep­tical of billion­aires. Espe­cially those suddenly claiming to care about the purchasing power of workers. Are they fibbers or just fools? I think it’s more subtle: these billion­aires are performing an idio­syn­cratic kind of class empathy, prob­ably at the behest of publi­cists. This feeling doesn’t come natu­rally. So, like a cyborg reading a love poem, it’s stilted. The persis­tent lack of detail signals that these billion­aires have no idea how, exactly, AI will accom­plish this. But the subtext is clear: If toler­ating worker purchasing power is what it takes to become a tril­lion­aire, that’s a sacri­fice I’m willing to make.

Some­times this theory goes by fancy names like pros­perity or abun­dance. Some­times it’s linked to specific policy mech­a­nisms like universal basic income. It all comes back to the same idea. I’m just going to call this stuff goodies. The theory is that under AI, citi­zens are going to get some nice level of goodies.

A few reasons we might be skep­tical of AI-goodies theory. The price of any consumer good has a lower bound set by the cost of its inputs (if one is not selling at a loss). But labor is only one input. So even if AI reduces labor costs to $0, the cost of other inputs—e.g., energy and raw mate­rials—remain. Even if AI lowers commodity prices by increasing their supply (say, through better discovery and extrac­tion tech­niques), they are still rival­rous resources. Not all are avail­able domes­ti­cally. Postwar, rising global consump­tion of oil in the West influ­enced the forma­tion of OPEC by future petrostates who sought to capture a share of oil-derived wealth. Simi­larly, in a post-AI US economy, foreign nations with crit­ical minerals—e.g., lithium, cobalt, nickel—may like­wise cartelize them, raising costs and creating a price floor. Related kinds of supply control may also arise domes­ti­cally. Today, a certain US agri­cul­ture company uses patented seeds to exert control over a market of osten­sibly inde­pen­dent farmers. If AI dramat­i­cally reduces the costs of manu­fac­turing and agri­cul­ture, Big AI may seek to capture value from the facto­ries and farms that benefit (say, through a sales royalty). AI-goodies theory implies that Big AI will allow maximum produc­tion of goods, leading to lowest prices. But the huge finan­cial projec­tions around AI—set by Big AI itself—may force them to do other­wise. That is, Big AI’s promises of world-histor­ical goodies for consumers and world-histor­ical profits for share­holders may be incom­pat­ible. If so—guess who prevails?

Still, to be fair to Big AI, I’ll accept their argu­ment that citi­zens will enjoy higher stan­dards of mate­rial well-being under AI through goodies. But the goodies must be paid for. Somehow, wealth must be trans­ferred from Big AI to everyone else. These trans­fers can happen by three basic methods:

1. Direct. Big AI directly provides goodies to citi­zens. As one AI CEO put it, to “capture much of the world’s wealth … and then redis­tribute this wealth to the people”.

2. Tax and spend. Govern­ment taxes Big AI at some high rate, and spends the money on public goodies. But the idea that Amer­ican AI compa­nies would consent to high tax rates to ensure public goodies is incon­ceiv­able because it vests more control in govern­ment, which no free-marke­teer believes should be so.‡

3. Market-based. Big AI dramat­i­cally reduces the cost of goods and services, which then flow into the market at extremely low prices. The flaw in this method is that AI will reduce prices of certain things more than others. For instance, it’s easy to see how AI will make streaming movies free; food and health­care and trans­porta­tion, less easy.

These methods aren’t mutu­ally exclu­sive. I expect the most likely method is the first, where Big AI directly provides the goodies, in the style of Gilded Age company towns. Big AI would avoid govern­ment inter­ven­tion through taxa­tion and redis­tri­b­u­tion. Big AI would control the goodies and who gets them. Big AI would be clearly under­stood as the goodie provider. Assuming the massive profits mate­ri­alize, Big AI could afford to make diverse baskets of goodies that include even things like food and health­care and trans­porta­tion.

Keep in mind the goodies economy will play out against a back­drop of wors­ening AI-induced labor-market effects—high unem­ploy­ment, but also wage declines, under­em­ploy­ment, misem­ploy­ment, and so forth. There­fore, a side effect of the goodies economy is that what­ever we think consti­tutes the social safety net will shift from a service of govern­ment to a service of private industry, specif­i­cally Big AI. One AI CEO shame­lessly predicts exactly this outcome: that there is “oppor­tu­nity for AI to be used to help provi­sion govern­ment services … includ[ing] health services, the DMV, taxes, social secu­rity, building code enforce­ment and so on.” But don’t worry—he concludes that because “poorly imple­mented services” cause “cyni­cism about govern­ment”, AI inter­me­di­a­tion would “strengthen[] respect for demo­c­ratic gover­nance.” And to be clear—by “AI”, he means “my privately owned AI”, and by “oppor­tu­nity … to help” he means a massive govern­ment contract.

This portends a polit­ical shift. Citi­zens once treated govern­ment as a polit­ical bulwark against the encroach­ments and over­reaches of industry. In the goodies economy, citi­zens will tend to align with Big AI, since it provides the goodies, which for many citi­zens may include basic needs. Under those condi­tions, if Big AI has a conflict with govern­ment—whose side will citi­zens take?

Indus­tries have often insu­lated them­selves against polit­ical conse­quences with popular prod­ucts. Arguably, in the last 25 years, the tech industry—espe­cially smart­phone and social-media compa­nies—has bene­fited from light regu­la­tory scrutiny rela­tive to its size. Exam­ples from the last century include the auto­mo­tive and fast-food indus­tries, which like­wise enjoyed decades of gentle regu­la­tion, even though both had serious public-health impacts.

To improve public accep­tance, Big AI has so far empha­sized their fun consumer-facing prod­ucts. These prod­ucts—e.g., gener­ating an image, or writing a story, or riding in a driver­less car—make AI seem harm­less: the citizen thinks of their own usage as trivial and fun, and there­fore models the aggre­gate effect of AI as a sum of simi­larly trivial indi­vidual usages. Consis­tent with this, Big AI has often framed these tools as “democ­ra­tizing” (ex. 2, ex. 3) implying that oppo­nents of these systems are in some sense anti-democ­racy (or as Winner predicted, antitech­nology and antiprogress).

In the goodies economy, as citi­zens lose inde­pen­dence as economic actors, they will also lose inde­pen­dence as polit­ical actors. They will be captured by industry. Econ­o­mist Yanis Varo­ufakis has called the current early stage of this arrange­ment “tech­nofeu­dalism”, a term suggesting how late capi­talism erodes indi­vidual economic agency. Through its provi­sion of goodies, Big AI intends to complete this tran­si­tion.

An economic histo­rian might say that a polit­ical takeover by Big AI is unlikely, citing successful worker-led polit­ical move­ments in the US and Europe. If the US reaches 30% AI-induced unem­ploy­ment, the thinking goes, workers will rise up and demand change through polit­ical mech­a­nisms of our liberal democ­racy. There’s reason to doubt these paral­lels. These events occurred earlier in the tech­no­log­ical era, when industry and workers were adver­sarial yet remained inter­de­pen­dent. Citi­zens retained irre­ducible leverage with both industry (as laborers) and govern­ment (as voters). In the decades after the US Gilded Age, this combi­na­tion of economic and elec­toral power was crit­ical to the rising status of workers. But in a future post-AI economy, workers could lose their jobs because of Big AI while remaining depen­dent on Big AI for basic needs. How do you bite the hand that feeds? Further­more, many previous conflicts between industry and workers were not gentle adjust­ments, but violent (say, the Luddite move­ment of the 1810s, and Gilded Age strikes of the late 1800s) or caused devas­tating, perma­nent social tran­si­tions (say, the 1917 Russian Revo­lu­tion).

Legally, shifting the social safety net from govern­ment to Big AI also means that citizen enti­tle­ment to these goodies will shift from public statu­tory and consti­tu­tional law to private-contract law. Private compa­nies can with­hold goodies for many reasons the govern­ment cannot. For instance: if you talk shit about the govern­ment, you will still get your unem­ploy­ment check, because the First Amend­ment says so. But no such restric­tion on Big AI: if you talk shit about them, they will be free to revoke your goodies. Can’t happen? Already does: credit reports are a private system that largely controls who gets favor­able interest rates. Or employer-provided health­care, a struc­turally unnec­es­sary tradi­tion that distorts incen­tives for workers by making their continued health contin­gent on continued employ­ment. Most likely, the polit­ical align­ment between citi­zens and Big AI will not only be enforced by carrot, but also by stick.

Of course, an industry seeking to induce polit­ical align­ment from citi­zens would benefit not only from distrib­uting goodies, but also weak­ening the govern­ment. In 2023, would you have believed this predic­tion? That an AI CEO would endorse a pres­i­den­tial candi­date. That the candi­date would win. That on his first day in office, the new pres­i­dent would install the AI CEO in the govern­ment with broad powers, in pursuit of a nebu­lous project of cutting waste and saving money. That ulti­mately, this project would save little money and primarily impair govern­ment agen­cies designed to protect the vulner­able. That somehow the parts of the govern­ment that provided large contracts for the same AI CEO would remain untouched. To be fair, for centuries US busi­nesses have sought to bend US public policy to suit their inter­ests, often with great success. This is a theme of Galbraith’s book The Predator State. What’s new is the mous­tache-twirling flagrancy of the project, and that it was under­taken at the behest of the newly elected pres­i­dent. Some called it a coup, though one premised on abdi­ca­tion of power rather than seizure.

‡ As of June 2026, reporting has suggested that an AI CEO has proposed to the current US pres­i­dent that the US take an equity stake in the AI CEO’s company (and maybe others). This arrange­ment would be very different from taxa­tion. It fore­sees a transfer of capital from the US govern­ment to the corpo­ra­tion to buy equity. As a share­holder, the US govern­ment would be enti­tled to a share of future profits, which may never exist. It also poten­tially creates a conflict of interest: as a Big AI share­holder, would the US govt regu­late or legis­late the AI industry? Taxa­tion, by contrast, requires no prepay­ment by the US govern­ment, and can be assessed on the current activ­i­ties of the AI company, regard­less of prof­itability. Histor­i­cally, the US govern­ment has taken equity stakes in private corpo­ra­tions as an indi­rect bailout. Which is maybe the better word for what is being proposed here. Absent this exigency, taxa­tion is econom­i­cally and polit­i­cally prefer­able. This AI CEO may also want to ask his chatbot what happened to the oil industry in Venezuela under Chávez.

Know your enemy

Oh Karl, the world isn’t fair
It isn’t and never will be
They tried out your plan
It brought misery instead
If you’d seen how they worked it
You’d be glad you were dead.
—Randy Newman “The World Isn’t Fair”

To iden­tify as a Marxist today is rare (at least in the US). More often, it’s wielded as an anti-capi­talist epithet against someone who endorses socialist or Commu­nist methods. True—Marx himself supported these methods. After all, he did write The Commu­nist Mani­festo. But short­hand Marxism tends to omit the reasoning that Marx, a philoso­pher and histo­rian, consid­ered vital. After all, he did write Capital.

Marx saw class struggle as the key insti­gator of history. He argued that capi­talism was polit­i­cally unstable, and neces­sarily led to worker revolt. Marx’s theory that economic value orig­i­nates in labor over­laps concep­tu­ally with Locke’s labor theory of prop­erty. But to Marx, capi­talism sepa­rated—or using his preferred term, alien­ated—workers from the fruits of that labor.

In prac­tice, Marx’s theo­ries have had mixed success. But Marx’s obser­va­tions in Capital about the inter­ac­tion of workers and machines remain rele­vant, even though he was writing in the mid-1800s, when the pinnacle of tech­nology was textile machinery. Looking back on worker upris­ings since the 1600s—including the Luddite move­ment that famously protested those textile machines—Marx wrote:

It took both time and expe­ri­ence before the workpeople learnt to distin­guish between machinery and its employ­ment by capital, and to direct their attacks, not against the mate­rial instru­ments of produc­tion, but against the mode in which they are used.

By protesting the textile machines, the Luddites were missing the bigger picture. The Luddites (and earlier workers) were raging against the literal machine—the specific new tech­nology. But the tech­nology was a symptom, not a cause. Instead, workers needed to rage against the bigger, figu­ra­tive machine—the extrac­tive capi­talist system.

Marx’s obser­va­tion has a subtler impli­ca­tion too. New tech­nology often holds itself out as the starting point of a narra­tive: from now on, every­thing is different. When we consider the tech­nology alone, that narra­tive domi­nates. But when we zoom out and consider the histor­ical context, the new tech­nology becomes the current endpoint of a much longer polit­ical narra­tive.

What would Marx say to AI critics—social, legal, economic, polit­ical—that have arisen so far? Maybe that we’re missing the bigger picture. That as a human inven­tion, AI may be the starting point of a new tech­no­log­ical narra­tive. But as an affront to human workers, it continues a long tradi­tion of capi­talist tech­nolo­gies, begin­ning with the Indus­trial Revo­lu­tion (if not earlier).

For their part, the capi­tal­ists who own a new tech­nology usually prefer to frame it as a starting point. Conve­niently, this omits the grubby histor­ical context that might dampen the marketing. Similar sleight of hand animates the snide techie’s favorite sick burn—Luddite—connoting a block­head irra­tionally opposed to tech­nology. Of course, this usage is delib­er­ately ahis­tor­ical—read Brian Merchant’s Blood in the Machine. Perhaps there should be a companion term for a block­head irra­tionally committed to AI.

These two frames for under­standing new tech­nology under capi­talism also suggest two corre­sponding analytic approaches to AI risk:

1. AI as tech­nology. On this view, AI is a starting point, and we consider the risks of what it can do that previous tech­nolo­gies could not. This isn’t wrong, exactly. It works for certain char­ac­ter­is­tics. For instance: I can believe that AI will facil­i­tate hacking and similar crim­i­nality like no tech­nology before. That risk is specific to AI as new tech­nology and deserves consid­er­a­tion.

2. AI as capi­talist instru­ment. But for other char­ac­ter­is­tics—say, AI’s effects on labor, public wealth, and the economy—thinking of AI strictly as new tech­nology embeds the error Marx warns against. For those char­ac­ter­is­tics, we should consider how AI may amplify or accel­erate existing trends. Put differ­ently—the ways in which AI is inher­ently polit­ical tech­nology.

In sum—as new tech­nology, AI acts as a creator of new risks; as a capi­talist instru­ment, it acts as an ampli­fier or accel­er­ator of existing risks and trends in the capi­talist system.

The strongest version of this framing is that the “ampli­fier or accel­er­ator” part is precisely Why AI Exists. Unprece­dented billions have already been poured into AI. Why? Because it’s expected to deliver more profit and concen­trate more wealth than any previous tech­nology. But ampli­fying these trends also means ampli­fying the atten­dant polit­ical risks. Earlier I accepted the argu­ment that Big AI will deliver abun­dant goodies to citi­zens—which AI propo­nents consider one of the strongest points in its favor. But doing so will likely increase align­ment between Big AI and citi­zens, with nega­tive polit­ical conse­quences. As usual—no free lunch.

This shift in framing also shifts the eviden­tiary context. When we think about AI risk, we’re neces­sarily making guesses about the future. But when we frame AI in the narrow sense of new tech­nology, we’re primarily consid­ering a time­line that starts now. Whereas when we shift to thinking of AI as a capi­talist instru­ment, we’re consid­ering a time­line that starts centuries ago and has evolved contin­u­ously into the present. We can and should study those existing economic and polit­ical trends, because those will likely shape the future trajec­tory. Put differ­ently: AI may be new. But it’s not immune to history.

The poisoned chalice

I think the biggest chal­lenge to AI in this country is polit­ical unrest … Can you make more money? It’s all irrel­e­vant if the country blows up … If I were sitting here in private with my peers, I’d be telling them … the country could blow up polit­i­cally. And none of us are going to make any money when the country blows up.
—an AI-adja­cent CEO

Big AI’s goal of labor replace­ment has two dimen­sions some­times over­looked, although they are both hidden in plain sight. The first relates to AI’s effort to replace knowl­edge workers. The other pertains to the compet­i­tive effects of doing so.

AI replacing knowl­edge workers. Labor replace­ment is the mech­a­nism. Struc­turally, Big AI seeks to build a moat around US economic growth by normal­izing AI as a cogni­tive inter­me­diary where knowl­edge work—roughly, econom­i­cally valu­able thinking and creativity—will happen.

This maneuver is part of the automa­tion play­book, espe­cially for the tech industry. During the rise of desktop automa­tion in the 1980s and 1990s, tech compa­nies sought to build moats by control­ling the file formats, programs, and plat­forms where docu­ments were created. As the commer­cial internet ripened during the 2000s, tech compa­nies sought moats based on lever­aging network effects—espe­cially the unholy dyad of adver­tising and social media.

How Big AI plans to profit from this inter­me­di­a­tion is an open ques­tion. One AI company has suggested taking a cut of AI-assisted discov­eries. The logis­tics and legal­i­ties would be boggling. Details—what­ever. For now, AI compa­nies largely agree on the first step: make workers depen­dent on AI to do their jobs, just as tech fore­bears made workers depen­dent on a certain soft­ware program to share a file, or on a certain website to have friends. This time, however, the soft­ware ulti­mately consumes the worker.

Big AI’s timing is canny. In recent decades, intan­gible assets have played an increasing role in global economic growth. An intan­gible asset lacks phys­ical embod­i­ment, and includes formal intel­lec­tual-prop­erty assets (e.g., copy­rights, patents) but also informal (e.g., brand good­will, other knowl­edge-work assets). Tangible assets are the oppo­site—bricks, mortar, inven­tory, and so forth. According to the WIPO, since 1995 global invest­ment in intan­gible assets has steadily increased, and since 2009 has exceeded that of tangible assets. By wide margin, the US leads invest­ment in intan­gible assets.

For a long time, intan­gible assets were under­valued because tradi­tional GDP measures and accounting prac­tices excluded them. In the 1970s and 1980s, during the first big wave of IT invest­ment, the US expe­ri­enced a coun­ter­in­tu­itive slow­down in produc­tivity growth, an effect econ­o­mist Erik Bryn­jolf­sson called the “produc­tivity paradox”. Part of the reso­lu­tion to the produc­tivity paradox is that returns from IT invest­ment were mani­festing not as labor produc­tivity per se, but as intan­gible assets. For busi­nesses, IT invest­ment was worth­while even if it didn’t increase labor produc­tivity, because it still gener­ated intan­gible assets. (Another part of the reso­lu­tion to the produc­tivity paradox is Bryn­jolf­sson’s theory of lagging comple­men­tary invest­ment, which will also likely apply to AI.)

Never­the­less, increasing labor produc­tivity matters a lot to workers. Why? It has histor­i­cally been the biggest driver of wage growth. Which brings us to another—maybe the most—conse­quen­tial trend in the US economy over the last 50 years: wage stag­na­tion, or the decou­pling of produc­tivity and wage growth. As the story is typi­cally told: from 1948 to 1973, wage growth (91.3%) basi­cally matched produc­tivity growth (96.7%). Since 1973, however, overall produc­tivity has grown (74.4%) much faster than wages (9.2%). Where has the money gone instead? Into the pockets of busi­ness owners, mostly. Over the same time­frame, the oper­ating profits of busi­nesses (as measured by net oper­ating surplus) has steadily increased. Today, if wage growth had kept up with produc­tivity gains since 1973, we’d expect wages to be more than double.

Together, these trends have been a long-term double whammy for US workers. Because of wage stag­na­tion, workers have been gener­ating increasing value for employers (as produc­tivity) but not capturing via wages the same share of that value they once did. Because of the shift toward intan­gible assets, employers have invested less in labor produc­tivity, further inhibiting wage growth. As AI replaces knowl­edge workers, we can expect it to amplify both trends.

Will it? Recent labor-market research by one major AI company suggests that the greatest “theo­ret­ical capa­bility” for AI labor replace­ment lies with tradi­tional cate­gories of knowl­edge work: finance, program­ming, engi­neering, science, legal, and the arts. A research group at Tufts Univer­sity was blunter: “the more AI helps you do your job, the more expend­able you can become. Finance profes­sionals, teachers/profes­sors, creative profes­sionals, accoun­tants and audi­tors, legal profes­sionals” are at risk. Even Bryn­jolf­sson said he was “surprised” that knowl­edge work would be so readily replaced.

“Tech­nology always makes certain jobs obso­lete; new ones will arise.” AI’s predicted labor replace­ment is unprece­dented in three ways: the diver­sity of tasks replaced; its outsize effect on highly educated workers; and the back­drop of 50 years of wage stag­na­tion. Automa­tion-driven tran­si­tions aren’t neces­sarily easy, even when they’re narrow and the economy can absorb the workers. Those who hand­wave over the details should study histor­ical exam­ples. When you tell a large group of workers that their skills no longer have economic value, you risk a polit­ical and social tinderbox. Recall Carl Benedikt Frey’s comment: “the short run can be a life­time”.

Confi­den­tial to owners of capital: nothing I’m saying above depends on a norma­tive view about wealth distri­bu­tion. I person­ally oppose wage stag­na­tion and believe workers should enjoy wage growth propor­tional (at minimum) to produc­tivity growth. If you don’t—fine. The argu­ment persists. Even accepting that the AI labor tran­si­tion is econom­i­cally rational and neces­sary, it will still likely have polit­i­cally desta­bi­lizing effects that you, owners of capital, will find unpleasant. Plus, you’ll get punched in the wallet too—keep reading.

The compet­i­tive effects of replacing knowl­edge workers. The previous section consid­ered how AI labor replace­ment will affect the rela­tion­ship between workers and employers. But it will also induce more carniv­o­rous compe­ti­tion among firms that adopt AI, and thus indi­rectly among the owners of capital. Whereas the last 50 years of the US economy have featured wage stag­na­tion, the post-AI economy may feature some­thing akin to capital stag­na­tion, for non-AI cate­gories of capital. A growing economy used to be a rising tide that lifted all boats. Post-AI, it may only lift a handful of yachts. Everyone else will be in dry dock. Arguably that capital tran­si­tion is already underway.

How might this work? The economy is increas­ingly driven by intan­gible assets. Knowl­edge workers provide those assets. If a company lays off its knowl­edge workers in favor of adopting some vendor’s AI, at first it gets the same produc­tivity at much lower cost. So much win.

But in doing so, the company commodi­tizes its own output. If your company can auto­mate its output via AI, others can too. Your existing competi­tors, certainly. But also upstarts who don’t have your cost foot­print. What­ever intan­gible assets AI can generate will be produced in excess, leading to a defla­tionary market for that asset. A company’s knowl­edge workers may be its greatest expense. But they also contribute to its compet­i­tive moat.

Further­more, the ongoing market value of a company is neces­sarily tied to its differ­en­ti­ated assets. A company valued at a billion dollars that adopts AI will then have to build some new asset not replic­able with AI, or it won’t be worth a billion much longer. Every­thing that makes AI an excel­lent cost collapser makes it an equally excel­lent capital collapser. A poisoned chalice. The greatest irony will be if AI-adopting compa­nies—having laid off knowl­edge workers en masse—hire them back merely to ensure that competi­tors cannot. In recent tech history, there have already been signs of this dynamic.

Consider large law firms, aka Big Law. Currently certain legal-AI star­tups license LLMs from Big AI and repackage them for Big Law at high prices. These star­tups claim to add other special sauce. OK, sure. Where’s the economic equi­lib­rium? If legal-AI star­tups prove that money can be made selling AI to Big Law—won’t Big AI just sell to Big Law directly, and cut out the star­tups? Or if legal-AI star­tups prove that AI can effec­tively provide legal services—won’t legal-AI star­tups just sell to clients directly, and cut out Big Law? Won’t members of Big Law that adopt AI have to lay off a lot of equity part­ners, because adop­tion of AI will shrink profit margins? Won’t the members of Big Law refusing AI have to consol­i­date to preserve their margins? Or just cave to AI? (The tendency of competi­tors to adopt similar prac­tices is called insti­tu­tional isomor­phism.) So it goes. Most states prevent nonlawyers from sharing in legal fees, so law firms will prob­ably remain a distinct set of enti­ties. But one plau­sible equi­lib­rium is that legal-AI star­tups disap­pear (quickly), and members of Big Law consol­i­date (rela­tively quickly) until there are only a handful left, all contracting directly with Big AI.

Along these lines, I expect that to succeed finan­cially, Big AI will likely need to demolish a signif­i­cant number of existing tech compa­nies and grab their revenue for itself. By the process described above: Big AI essen­tially uses its tech customers as an R&D facility. Big AI licenses models to these compa­nies. Tech compa­nies compete to adapt their busi­nesses to AI. Once a concept is proven, Big AI directly takes over that market. The labor-replace­ment story will grow into a company-replace­ment story. Many of those tech compa­nies—and their share­holders in the public markets—may also find that AI is a poisoned chalice.

The resource curse

The theory is thought also to fit Middle Eastern oil exporters espe­cially well. In this region, govern­ments’ access to rents, in the form of oil revenue, may have freed them from the need for taxa­tion of their peoples, and that this in turn freed them from the need for democ­racy.
—Jeffrey Frankel

Since AI is new, there are not yet exam­ples of post-AI economies. So we might ask a broader ques­tion: what happens to nations whose economies are concen­trated around one inher­ently polit­ical tech­nology? The most instruc­tive compa­ra­bles in the world economy are prob­ably petrostates: nations whose national wealth is concen­trated in oil or natural gas. The analogy is imper­fect. Unlike AI, oil is anchored to geog­raphy and limited in supply. On the other hand, oil has proven finan­cial value; AI does not.

Petrostates are vulner­able to what econ­o­mists call the resource curse, with two distinc­tive sets of effects:

1. Capital effects. The concen­trated resource tends to pull labor and capital from the rest of the economy, depriving other economic sectors, espe­cially manu­fac­turing and agri­cul­ture. This makes the national economy more depen­dent on imports for basic needs, and more sensi­tive to price volatility of the resource. If the resource price declines suffi­ciently, the nation will be unable to keep affording imports, but also unable to fulfill those needs with domestic produc­tion.

2. Polit­ical effects. The value of the concen­trated resource creates what Jeffrey Frankel calls “a polit­ical contest to capture owner­ship”, which in turn encour­ages the emer­gence of auto­cratic or oligarchic insti­tu­tions captured by an economic elite who seek to retain control of the resource. The process is self-rein­forcing in two ways. First: the economic elite uses its wealth to repress polit­ical oppo­nents. Second: as the govern­ment derives more income from the concen­trated resource, it relies less on taxa­tion of citi­zens, which weakens demo­c­ratic account­ability.

According to the Council on Foreign Rela­tions, “petrostates include Algeria, Cameroon, Chad, Ecuador, Indonesia, Iran, Kaza­khstan, Libya, Mexico, Nigeria, Oman, Qatar, Russia, Saudi Arabia, the United Arab Emirates, and Venezuela.” No one would confuse this with a list of successful liberal democ­ra­cies. Due to its increasing oil and gas produc­tion, some also consider the US a petrostate.

Which petrostates provide the most goodies to citi­zens? The Gulf petrostates—Qatar, UAE, Saudi Arabia, Oman. These nations distribute petrowealth to citi­zens through substan­tial public bene­fits (e.g., housing, health­care, educa­tion, public-sector jobs) while having low rates of taxa­tion. The tradeoff, of course, is that these coun­tries are not even notion­ally demo­c­ratic. They are hered­i­tary monar­chies. They offer none of the bene­fits that citi­zens in a liberal democ­racy enjoy (or some­times take for granted)—demo­c­ratic partic­i­pa­tion, indi­vidual civil rights, freedom of speech.

But the Gulf states were never liberal democ­ra­cies. Can the resource curse send a liberal democ­racy spiraling into autoc­racy? Sure—consider Venezuela. Today, Venezuela is an author­i­tarian regime. But from 1958 to the 1990s, it was a stable liberal democ­racy. This changed with the 1998 elec­tion of Hugo Chávez. After taking office, Chávez pursued populist economic poli­cies funded by national oil wealth. Embold­ened by high oil prices, Chávez amended the consti­tu­tion to bolster his own power and installed loyal­ists at the top of Venezuela’s national oil company (PDVSA). He redi­rected oil revenues to a series of domestic social programs (the “Boli­varian missions”). Chávez lever­aged his resulting popu­larity among workers to consol­i­date his power, for instance by ending term limits and taking control of the judi­ciary. (Confi­den­tial to AI CEOs: think care­fully before you offer an owner­ship stake to any US pres­i­dent who has author­i­tarian and populist tenden­cies.) After oil prices dropped in 2014, the resource curse hit Venezuela hard, putting its economy into crisis and allowing Chávez’s successor Nicolás Maduro to estab­lish a dicta­tor­ship. Demon­strating a central irony of the resource curse: while the resource is gener­ating wealth and deliv­ering goodies to citi­zens, author­i­tarian moves are depicted as essen­tial to sustaining pros­perity. After the economy weakens, author­i­tarian moves are depicted as essen­tial to restoring pros­perity. The ratchet only tightens. The river flows only down­hill.

Norway is another outlier. Norway is a robust liberal democ­racy. After oil was discov­ered in its North Sea terri­to­rial waters in the 1960s, Norway created the state-owned Statoil to build and operate oil facil­i­ties. In 1990, Norway estab­lished the Govern­ment Pension Fund Global, a sover­eign-wealth fund to receive surplus oil revenues and invest them for the benefit of Norwe­gian citi­zens. According to Norges Bank, “[t]he aim of the fund is to ensure that we use this money respon­sibly, think long-term and so safe­guard the future of the Norwe­gian economy.” In one sense it has worked: the GPFG currently holds over $2 tril­lion in assets. What’s the problem? Even Norway hasn’t entirely avoided the resource curse. Rising sover­eign wealth has deliv­ered tons of goodies to Norwe­gians. But it has also made Norway’s govern­ment less finan­cially disci­plined and created social conse­quences like high dropout and unem­ploy­ment rates. (A popular 2025 Norwe­gian book was called The Country That Got Too Rich.) Keep in mind that Norway reached this predica­ment despite 50 years of careful prepa­ra­tion and over­sight. Inher­ently polit­ical tech­nolo­gies can upend even gener­a­tional prepa­ra­tions. Those who would suggest that the US can follow a similar model with AI: easier said than done.

Putting it all together: Among AI risks, we should take more seri­ously the poten­tial conse­quences of AI working as intended. AI is a capi­talist instru­ment. Its prin­cipal func­tion is to concen­trate capital. Its intended mech­a­nism is large-scale labor replace­ment. But it is also inher­ently polit­ical tech­nology. As AI makes it harder for workers to capture value from their labor, they will increas­ingly have to rely on goodies from Big AI, priva­tizing what were once func­tions of govern­ment. If Big AI subsumes the func­tions of workers and govern­ment, both will tend to realign polit­i­cally around Big AI’s inter­ests. What­ever term describes this system, it is not liberal democ­racy as US citi­zens have tradi­tion­ally under­stood it. AI-centered capi­talism risks an extinc­tion of demo­c­ratic possi­bility. It will be America. But it will no longer be Amer­ican.

Epilogue

Thank you for reading.

Every citizen has a voice in these issues. Every citizen can partic­i­pate in the vital public conver­sa­tion about how we want AI to be part of our country—our schools, our work, our fami­lies.

Joseph Weizen­baum: “The myth of tech­no­log­ical and polit­ical and social inevitability is a powerful tran­quil­izer of the conscience.” 

In the upcoming elec­tion, vote for people who take your rights and inter­ests seri­ously.

I support orga­nized labor.

Ulti­mately, Big AI is constrained by an incon­ve­nient truth. Today, they need us more than we need them. As long as that remains true, we the people have the upper hand.