19.06.26 Demonetized Sovereignty: Emerging Nations Between Neoliberalism and the Age of AI
Demonetized Sovereignty: Emerging Nations Between Neoliberalism and the Age of AI
*(Simplified English version — same essay, same ideas, easier sentences)*
Most writing on AI today treats it as a fight between states and companies — Washington against Anthropic, America against China, government power against corporate power. If we look at this fight from Patna instead of Washington, it looks the same on the surface but very different underneath. The political question — who will own the most powerful technology of this century — is real. But under this question sits an older one, which emerging nations have already faced before, in different forms: what happens to a people when the tools of their dependence arrive in the language of freedom. Railways once carried this same trick. So did structural adjustment programs from the IMF and World Bank. AI is only the newest carrier of an old cargo. To understand this properly, we must move from politics into philosophy — and from philosophy into a harder question: what is left of a person once productivity has finished measuring them.
## I. National Interest versus Corporate Capture
We usually assume that the nation-state has full control over force and decision-making inside its own borders. This was never fully true, but AI now exposes the gap clearly. A small number of private companies now hold something governments have never had to share before: the power to decide — alone, and based on business logic — who gets to think faster, diagnose illness better, or plan more accurately. The old idea of "national interest" — five-year plans, public-sector steel factories, the postcolonial dream of self-reliant industry — assumed that the state could at least see the machine it was trying to control. AI infrastructure takes away even that basic visibility, even from the governments that depend on it.
But corporate capture does not only come from outside the country. It also happens inside, and India's own welfare system shows this clearly. Aadhaar-linked biometric verification now controls access to rations, pensions, and rural job guarantees for hundreds of millions of people. By the government's own data, about six percent of the tens of millions of monthly verification attempts fail — and this rate has not improved in over ten years. Failures happen most often among manual laborers and elderly citizens, whose worn fingerprints and changed iris patterns no longer match the database. Santoshi Kumari, an eleven-year-old girl in Jharkhand, died of starvation after her family's ration card was cancelled because it was not linked to Aadhaar. A 2020 J-PAL study found that most of the ration cards cancelled in that state belonged to real, eligible families — not to "ghost" beneficiaries, as officials claimed. These are not small mistakes in an otherwise good system. This is what it looks like when a state runs its most basic duties — feeding people, paying pensions — through technology it did not design and cannot fully check. This is consumption capture happening one level above the citizen: the state itself becomes a consumer of governance technology, ruling its people through judgment it has rented rather than judgment it has earned.
To be fair, this capture has not gone without an answer. India's own response — the IndiaAI Mission, backed by about ten thousand crore rupees, and the sovereign language models shown by Sarvam AI and the IIT Bombay-led BharatGen group at the 2026 AI Impact Summit — is a real attempt to win back control of the architecture, instead of staying a permanent tenant of someone else's intelligence. Whether this becomes a true correction, or just a more advanced version of the same dependence — paid for by the nation, but still built on the same productivity logic — is the open question this essay tries to answer. The deeper danger, either way, is not that companies will openly disobey the national interest. It is that "national interest" will quietly start to mean only "whatever keeps the system running smoothly" — efficiency, uptime, compliance — instead of what it used to mean: building a population's own ability to govern itself.
## II. Developed and Emerging Nations: The New Architecture of Dependency
The gap between rich and poor nations in the AI era follows an old colonial trade pattern, almost exactly. Raw material is taken cheaply from the periphery (the poorer, outside regions), refined at the center (the rich core), and sold back to the periphery at a price it cannot produce on its own. The old economy took raw cotton and sent back finished cloth. The new economy takes behavioral data, language, and the cheap or unpaid labor of people who supply the "training signal" for AI — and sends back "intelligence" as a paid subscription.
The clearest example of this pattern is real, not a theory — it happened in Kenya. Starting in late 2021, OpenAI hired the outsourcing company Sama to have workers in Nairobi read and label some of the most disturbing text on the internet — descriptions of child abuse, violence, and self-harm. This work let OpenAI build a safety filter, so this content would not reach a paying user somewhere else in the world. The workers were paid between roughly one and two dollars an hour. Several later said they suffered lasting psychological harm, and some took legal action and signed a petition to Kenya's parliament asking for protection. ChatGPT's praised safety and the Kenyan workers' documented trauma are not two separate stories. They are the same transaction, seen from its two ends: value collects at the center, cost is absorbed at the periphery, and there is no system through which the periphery shares in the value it helped create.
This is the difference between operational sight (knowing how to use a tool) and architectural sight (knowing how the tool was built, and why). It echoes an old colonial education policy — the kind that trained clerks who could run the empire's paperwork, but never trained scientists who could redesign its instruments. A population can be given wide permission to use a system, and no permission at all to question why it was built that way. These two things are easy to confuse, especially for governments eager to look digitally modern.
But this picture is not fixed, and an honest account must also show the push-back, not only the extraction. Brazil's central bank built PIX, an instant payment system now used by more than three-quarters of Brazil's adults, almost entirely outside private global payment networks. Brazil's National AI Plan has also committed several billion dollars to its own LLM ecosystem, including the Portuguese-language model Sabiá and a sovereign cloud project — though Brazil's own planning papers openly admit that dependence on imported chips is still the biggest block to real independence. Indonesia is digitizing its social welfare system through a national digital ID linked to health and welfare data. India's sovereign AI models, mentioned earlier, are a third example of this same instinct. None of these efforts has yet closed the architectural gap. But all of them show that emerging nations feel this gap as a daily administrative fact, not as an abstract idea — and are responding to it. Scale — billions of users, hundreds of languages, the combined size of India, Indonesia, Brazil, Nigeria, and Kenya — is the resource most often named as the periphery's counter-weight, and it is real. But scale only becomes real leverage through deliberate institution-building like this. Left alone to market forces, scale instead becomes the largest possible captive market — and the cheapest possible labeling workforce.
## III. The Final Result of Neoliberal Policy: From Captured Institutions to a Captured Self
At this point, it is tempting to use the word "neoliberalism" loosely — as a general mood, a vague complaint about capitalism — rather than as a specific history of institutions. We should resist that temptation, because the more exact claim is also the more serious one. The particular shape AI has taken — privately owned, controlled by a few frontier labs, built to maximize engagement and productivity rather than designed as public infrastructure — is not an accident that happened to a neutral technology. It is the direct result of forty years of deliberate restructuring in exactly the sectors that decide where powerful new technology gets built, and what purpose it is built to serve.
Start with research itself. In the United States, the Bayh-Dole Act of 1980 let universities patent and license inventions that public money had paid for. This turned a shared research commons into a stock of private property almost overnight. Over the following decades, the center of advanced computing research moved away from publicly funded, mission-driven programs — the same line of work that, through DARPA funding, had produced the internet's basic protocols — and moved toward startups funded by venture capital, and the research divisions of a handful of trillion-dollar companies. By the time large language models became technically possible, the ability to build them at scale existed almost nowhere except inside a small number of private companies. This was not because public science had lost its ability to think of such ideas. It was because forty years of policy had steadily moved the computing power, the trained people, and the right to commercialize new ideas, away from public hands.
The same restructuring reached emerging nations through a different door: structural adjustment. Loan conditions set by the IMF and World Bank in the 1980s and 1990s forced many Global South governments — India and Brazil among them — to cut public spending on education and health, sell off state enterprises, and reorganize remaining public services around cost-recovery targets and performance metrics. This is why AI diagnostic tools and AI tutors now fit so smoothly into many emerging-economy health and education systems: the public alternative — a salaried doctor with unlimited time for a patient, a teacher answerable to a community rather than to a dashboard — had already been worn down before any algorithm arrived. The same pattern reaches the military world through a different route: the privatization of logistics, intelligence work, and now frontier-AI defense contracts to companies like Anduril and Palantir, with OpenAI itself entering defense deals in the past two years. This repeats, at the level of national security, the same shift of power from public, accountable institutions to private, profit-driven ones.
What connects research funding, education, health, and military spending is not chance. It is one single belief, repeated across forty years of policy: that public management is wasteful, and private incentive creates value. AI is the field where this belief reached its purest form, because AI's whole promise — efficiency, optimization, removing friction — is the technological version of the same logic that hollowed out the public sector elsewhere. This is why the blame here belongs specifically to neoliberalism, and not to "capitalism" or "technology" or "modernity" in some vague general sense. Neoliberalism names a specific, identifiable project — with a starting point, known authors, and a clear list of sectors it deliberately rebuilt. AI is not separate from that project. It is its most advanced product.
The same logic, applied to a person instead of an institution, produces thinklessness. Neoliberalism's real achievement was never just privatization. It was turning the self into a unit of measurable productivity — what Foucault called *homo economicus*, and what Byung-Chul Han calls the achievement-subject: a self that exploits itself while believing it is exercising freedom. AI does not break this logic. It completes it, by offering exactly the technology this logic was waiting for: a way to outsource not just labor, but judgment itself, while still keeping the appearance of personal choice. Once judgment is rented instead of exercised, citizenship shrinks to what I have called elsewhere *biological citizenship* — the self reduced to its status as an eligible or ineligible recipient, a verified or flagged entry in someone else's database. Arendt's distinction between labor, work, and action shows us exactly what is lost first. Labor — keeping the body alive — survives an AI economy easily, and may even become easier under it. Action — the ability to stand before other people and start something genuinely new in public life — withers away, because action needs exactly the friction that AI's whole business model exists to remove. The institutional capture described above, and the captured self described here, are not two separate arguments. They are the same argument, seen at two different sizes: the hollowing of public responsibility at the level of the nation, and the hollowing of judgment at the level of the person. The second is simply what the first feels like from the inside.
## A Bridge: Why Witness-Consciousness Is the Right Answer Here
If the capture happens at the level of being — *ontological* capture — the answer cannot simply be a better set of numbers: a more humane productivity index, a gentler algorithm, a kinder dashboard. Any answer like that still works inside the same logic it claims to oppose. It argues about what should be measured, without questioning the right of measurement itself to rule over us. And a clever enough system will eventually learn to optimize for "humane" too. What we need is not a different claim about what a person's output should look like. We need proof that some part of the self is not output at all — not because we have not yet built the right tool to measure it, but because witnessing and measuring are simply two different kinds of activity, from the start.
This is exactly how the Upanishadic idea of witness-consciousness answers ontological capture, rather than just comforting us against it. The *aham asmi* — "I am" — observes experience without being the same thing as any of its productive states. It is not a better-performing self. It is a part that is structurally left over, a remainder that no act of optimization can ever reach, because it is defined as the watcher of optimization, not as one more thing optimization could improve. Neoliberal AI capture works by making the self exactly equal to its measurable output. Witness-consciousness names precisely the part of the self that this equation can never finish capturing. It is not a hiding place from the productivity question. It is a quiet refusal to accept that question's right to judge us at all.
## IV. Returning to the Basics of Being Human: A Possible Way Out
### Why this resource, and not another
A reader who knows the wider Indian and South Asian intellectual tradition will rightly ask: why use witness-consciousness from the Upanishads as the main philosophical anchor here, instead of Buddhism, Gandhi, Tagore, Ambedkar, or Aurobindo? Each of these offers real resources for resisting exactly this kind of capture. The honest answer is that they are not really competing with each other. They are tools tuned to different jobs, and the choice made here is deliberate, not because the others are wrong.
**Buddhism's** idea of *anatta*, "no fixed self," is in some ways an even more radical answer than witness-consciousness. If there is no self to begin with, there is nothing for capture to grab onto. But this radical idea has a practical risk that we have already seen happen. Buddhist-style techniques — mindfulness, non-attachment, present-moment awareness — have been fully absorbed into corporate wellness culture over the last fifteen years. They have been repackaged as productivity tools that help employees tolerate heavier workloads rather than question them — a pattern critics call "McMindfulness." Any tradition built around a technique you perform is open to exactly this kind of re-absorption, because capture is very good at swallowing practices and turning them into products. Witness-consciousness, as used in this essay, is not a technique to perform on a schedule. It is a claim about what is already, and always, true about being — which makes it much harder to package as a daily habit or sell as an app.
**Gandhi's** ideas of *swaraj* (self-rule) and *satyagraha* (truth-force) work at the level of political action and bodily discipline. They answer the question of what one should do, and they are already present in this essay's focus on the panchayat and the public square as places of resistance. But Gandhi's framework assumes a self that is already capable of disciplined choice. It does not explain, on its own, what is left in a person whose judgment is quietly being captured before any choice is even made. Gandhi's ideas and the argument made here support each other rather than compete. One supplies the politics of the answer. The other supplies its foundation in being.
**Tagore's** criticism of the mechanical, calculating "Nation," and his defense of personality and creative freedom against organized civilization, is closest in spirit to the argument made here. It works at the level of civilization and art, rather than strict philosophy. He is less an alternative to this essay's claim than its cultural twin — his distrust of mechanism over personality could easily have supplied this essay's opening quote.
**Aurobindo** is really the closest ally here, not an alternative. His whole system of integral yoga is built directly on the Upanishadic witness, and he adds an evolutionary dimension to it — the claim that recovering this layer of consciousness is not going backward into a pre-colonial past, but joining an unfinished historical task. This essay's answer should be read in that same spirit: not nostalgia, but an unfinished project picked up again.
**Ambedkar** belongs on this list for a different, more demanding reason than the others. Leaving him out would be a far more serious mistake than choosing the Upanishads over any one rival tradition. Ambedkar's method was critical retrieval — never accepting a tradition wholesale, just because it is inherited. He broke from Hinduism precisely because he judged that its social order could not be separated from its texts. He also rejected Gandhi's romantic picture of the village republic, calling it a den of narrow localism and caste enclosure rather than a school of liberty. To use Upanishadic witness-consciousness honestly — in Ambedkar's spirit, not against it — we must clearly set aside that same textual tradition's history of Brahminical hierarchy and caste exclusion. The philosophical idea can travel on its own. The social order in which it was first written down cannot, and does not need to come along with it. What this essay borrows from Ambedkar is not a rival metaphysics, but this discipline of picking carefully what to keep. And it is the same discipline the next section needs to apply to the institutions it goes on to recommend.
### Building the way out, institutionally
This discipline is necessary, because the institutional answer to capture risks exactly the kind of romanticizing it should be guarding against. Arendt's distinction between labor, work, and action suggests rebuilding small, real spaces where action can still happen — the gram sabha, the panchayat meeting, the railway platform conversation — places where judgment is still used together, in person, and where AI's suggestions can be discussed, challenged, and tamed rather than simply accepted. Sen's capability approach offers the matching policy discipline: AI use should be judged not by *niti* — whether the rule was followed efficiently — but by *nyay* — whether the real freedom of the person in front of the system actually grew. India's Digital Public Infrastructure, and similar efforts visible in Brazil's PIX and Indonesia's digital ID rollout, point toward a third path between full corporate capture and full state control. But each of these, as their own planners admit, is still just a vessel — it carries whatever logic is poured into it.
But this essay has already used the panchayat and the railway platform twice as places of friction and good deliberation. Honesty requires us to also say what these places have been, and often still are. Ambedkar's own judgment of the village — not a republic of brotherhood, but, in his words, a structure that locks in and repeats caste hierarchy — must be taken seriously here, not set aside just to make a tidy ending. Panchayats have been tools of domination as often as they have been forums for deliberation. Oral tradition passes down patriarchy and exclusion just as easily as it passes down wisdom. "Waiting" is not a calm, thoughtful discipline for the woman kept outside a decision that men in her own household are making, or for the person standing for hours in front of a patwari's desk. It is the experience of being made small — a way the powerful remind the powerless of their dependence. "Ambiguity" is also not always depth. It is often the bureaucratic discretion that allows harassment and bribe-seeking — exactly what turns Bihar's ration shops and block offices into sites of extraction, not just sites of philosophy. A reader who has actually seen any of this happen would be right to reject an ending that simply asks us to return to these forms as they have really existed.
The answer argued for here, then, is not a return to these institutions exactly as they have existed. It is the deliberate pulling-out of specific useful features from them — friction, real physical presence, the need to persuade another human being rather than satisfy an algorithm, the fact that resolution cannot happen instantly — while clearly rejecting the hierarchy that has historically come attached to those same forms. This is harder than nostalgia, because it is a design problem, not a restoration project. We must build new spaces that keep the slowness and human presence that made the old spaces valuable, without inheriting the exclusions that made many of those same spaces instruments of harm. Emerging nations' deepest resource in the AI era may not be computing power at all. It may be exactly this: populations who still remember — however unevenly, and at however high a historical cost — what real deliberation among present, accountable, embodied people looks like. And who therefore have the standing to insist that whatever replaces it should be held to no lower a standard.
## Closing
The political question this essay began with — who owns the architecture of intelligence — turns out, on closer look, to be inseparable from an older question. Emerging nations have already answered a version of this question before: under colonial rule, under structural adjustment, and now under algorithmic governance. The question is whether independence won at the level of the flag survives at the level of the self. The institutional history traced in this essay suggests it does not survive automatically. Forty years of neoliberal restructuring across research, education, health, and defense created exactly the conditions under which AI now arrives already coded for extraction rather than for capability. The way forward is not just catching up with the technology, not just a sovereign LLM, and not a sentimental return to village India. It is the harder work of building new institutions that hold two things at once: the real conditions needed for human deliberation, and an honest, critical eye about which inherited forms truly deserve to survive into them.
Comments
Post a Comment