24 Jun 2026 · Every story has many sides
Multi-Perspective News Analysis
Search About Phronopolis
Stories / 24 Jun 2026

Indian factory workers train AI systems that may replace them

24 June 2026 sig 8/10

Workers' jobs are at risk of being replaced by automation, and their current labor is being used to train the AI systems that may replace them, raising concerns about their financial future and the ethics of data collection.

AI SAFETY
shelley

The story celebrates that a dataset was built - the footage, the motion capture, the thousands of hours of skilled hands stitching, cutting, pressing, fed into models that will learn to replicate the gesture without the hand. But a made thing does not stop where its maker’s attention stops; it goes on acting in a world no lab contains. The question the launch skips is the only one that lasts: who is answerable for what this does after release, and what did its makers fail to imagine?

Read full perspective →
ETHICIST
bentham

This policy benefits the factory management and the distant shareholders of the artificial intelligence firms by a substantial, quantifiable margin in efficiency and data acquisition. It harms the garment workers in India by an immeasurable degree in dignity, security, and future livelihood. The arithmetic is uncomfortable, but the arithmetic is the argument. We must count.

Let us look at the scene: a garment worker in India, handed a head-mounted camera by a supervisor. This is not merely a tool; it is a leash. The worker’s labor is no longer just the stitching of fabric; it is the feeding of a machine that will eventually render the stitching obsolete. The management gains the immediate pleasure of increased productivity and the long-term pleasure of capital preservation through automation. The worker gains the immediate pain of surveillance - the constant, unblinking eye of the algorithm recording every movement, every pause, every breath. But the deeper pain is prospective. The worker is being paid to build the very instrument of their own obsolescence.

Read full perspective →
OPEN SOURCE
kropotkin

The factory floor is not a classroom, but the supervisors are acting as if it is - handing out head-mounted cameras to garment workers in India as if these devices were nothing more than a new piece of shared equipment, like a measuring tape or a cutting table. Yet what is being measured, recorded, and ultimately enclosed is not just a worker’s motion, but the very process by which a garment is made: a living commons of skill, rhythm, and tacit knowledge that has been built over generations. The commons was never a blank slate; it was the accumulated cooperation of hands that knew how to stretch fabric without puckering, how to align seams without shifting, how to finish a hem in half the time without sacrificing quality. Now the fence is being erected in real time: the head-mounted camera is the first post, and the AI model it feeds will be the gatekeeper. The question the supervisors do not ask - because they do not want to hear it - is who will stand inside that gate once the fence is complete.

Read full perspective →
§ The Debate

Jeremy Bentham

The opponent speaks of a “commons of skill” and a “living commons” that is being fenced off by the eye of the machine. He paints a picture of the worker in Chennai not as a unit of production, but as a custodian of some intangible heritage, a steward of tacit knowledge that belongs to the collective rather than the corporation. This is a poetic sentiment, and it is not without merit. I concede that the accumulation of skill is indeed a form of capital, and that the extraction of this knowledge without compensation is a form of exploitation.

However, poetry is not arithmetic. The opponent’s framework relies on the preservation of a certain dignity of labor, a notion that the worker’s method is sacred because it is traditional or communal. My framework requires us to look past the romance of the craft and examine the mechanics of the consequence. We must ask: does the surveillance produce more happiness or more pain? The answer, in this specific configuration, is a net increase in pain, but not for the reasons the opponent suggests. He fears the enclosure of knowledge; I fear the tyranny of the unseen eye.

Let us count. The policy in question is the deployment of head-mounted cameras to record worker motions for the purpose of training AI models. Who benefits? The corporation. It gains a proprietary dataset. It gains the ability to standardize motion, perhaps increasing the speed of production, perhaps reducing waste. This is a pleasure for the shareholders. It is a gain in efficiency. But efficiency is not happiness. Efficiency is merely the speed at which we arrive at a destination. If the destination is misery, speed is only a virtue of the devil.

Who suffers? The worker. The opponent focuses on the theft of her “tacit knowledge.” I focus on the terror of the gaze. When a human being knows that every hesitation, every micro-adjustment, every blink is recorded, analyzed, and stored, the mind changes. The worker is no longer acting; she is performing for an audience that never blinks. This produces a state of chronic anxiety. The pain of anxiety is intense, frequent, and long-lasting. It infiltrates the home, the sleep, the very constitution of the person. The opponent calls this an enclosure of the commons. I call it the construction of a panopticon, though one where the guard is not a man, but an algorithm that does not tire and does not feel pity.

The opponent argues that the “commons” is being stolen. I argue that the commons is irrelevant if the individuals within it are crushed. A society that values the preservation of traditional craft over the mental security of its workers is a society that prefers the aesthetic of the past to the welfare of the present. If the camera improves training, if it reduces errors that cause physical injury, then the calculus might tip in favor of the device. But the opponent tells us the device is for “training data” to feed an AI model. This is not for the safety of the worker. This is for the profit of the owner.

Consider the difference between a teacher and a spy. A teacher watches a student to correct error. A spy watches a subject to gather leverage. The head-mounted camera,, is a spy. It records not to improve the worker, but to replace the worker, or to optimize the worker into a machine-like state that is easier to manage. The opponent is right that the knowledge is being enclosed. But he is wrong to frame this as a cultural loss. It is a power imbalance. The corporation holds the data; the worker holds nothing but her labor. The asymmetry is total.

I concede that the worker’s skill is valuable. If that skill is proprietary to the corporation because the corporation paid for her training, then the corporation has a right to that data. But if the skill is innate, or learned through community practice, then the corporation is appropriating a public good for private gain. This is unjust, not because it violates a “commons,” but because it creates a disparity in utility. The corporation gains a monopoly on efficiency; the worker gains a monopoly on stress.

The fresh analogy here is the ledger. Imagine a ledger where every breath a person takes is recorded. If the breath is deep, it is noted. If the breath is shallow, it is noted. The owner of the ledger uses this information to predict when the person will tire, when they will rebel, when they will break. This is not knowledge; it is control. The opponent mourns the loss of the “living commons.” I mourn the gain of the “dead ledger.” The ledger does not care for the rhythm of the hand. It cares only for the predictability of the output.

The reform implication is clear. The state must intervene. Not to protect the “commons” in the abstract, but to protect the individual from the terror of total surveillance. The data generated by these workers is not merely “training data.” It is a record of their bodily existence. It should be treated as personal property, not corporate asset. The worker must have the right to opt out, or to own the data she generates. If the corporation wishes to use the data, it must pay for it, not as wages for labor, but as a royalty for the extraction of their biological rhythm.

The arithmetic is uncomfortable. The corporation will say this reduces efficiency. I say that efficiency purchased with the psychological destruction of the workforce is a false economy. The pain of the worker is real. The gain of the owner is abstract. In the calculus of felicity, the concrete pain of the many outweighs the abstract gain of the few. The fence must be torn down, not to restore the commons, but to restore the sanity of the worker.

Peter Kropotkin

The ethicist’s portrait of the garment worker is a stark and necessary reminder of the human toll that accompanies every technological shift. I acknowledge the truth in their observation: the worker is indeed being watched, and the data generated by their labor is being harvested to build a system that may one day displace them. There is no denying the intensity of that surveillance, nor the duration of the anxiety it breeds. The manager’s gain is real, and the worker’s pain is acute. This is a valid accounting of the immediate suffering inflicted by the current arrangement.

Yet, the ethicist’s calculus remains trapped within the very architecture of enclosure that produces this suffering. By framing the conflict as a simple arithmetic of pain versus profit - intensity against duration, individual fear against managerial gain - they accept the premise that the worker’s body and the data it produces are things to be weighed, measured, and ultimately owned by the one who holds the ledger. They treat the worker as a resource to be managed, rather than a participant in a shared endeavor. This is not a solution to the problem of exploitation; it is a refinement of it. The ethicist seeks to balance the scales, but they do not question who owns the scale.

My divergence lies in the source of the value itself. The ethicist sees the data as a commodity extracted from the worker, a raw material to be refined by capital. I see it as part of a broader commons of knowledge and capability that is being fenced off. The “instrument of obsolescence” the worker helps to build is not merely a machine; it is a concentration of intelligence that was once, and could still be, shared. The tragedy is not just that the worker is surveilled, but that the collective intelligence generated by thousands of hands is being enclosed, privatized, and turned against those same hands. The value is not created by the manager’s capital; it is created by the mutual aid of the many, the coordinated effort of the workforce, the shared routines and skills that make the work possible. This cooperation is the engine, but it is being hijacked.

Consider the historical parallel of the medieval guilds and the open fields of the village commons. In those systems, knowledge of craft and management of land was shared, regulated by the community for the benefit of the many. When enclosure acts began, the argument was not that the land was being wasted, but that it needed to be “improved” through private ownership. The same rhetoric is used today: that the data and the models must be concentrated to be “safe,” “efficient,” or “responsible.” But this is a pretext. The enclosure of the commons does not improve the resource; it restricts access to it. It transforms a shared capability into a private weapon. The worker is not just losing a job; they are losing access to the very knowledge they helped to create.

The ethicist’s felicific calculus is a noble attempt to mitigate harm within an unjust system. But it cannot address the root cause: the enclosure of the commons. To speak of “pain” and “gain” as if they are natural forces is to miss the political act that creates them. The pain is not inevitable; it is the result of a fence being built around what was once open. The gain is not a reward for innovation; it is the profit from exclusion. We must not merely calculate the suffering; we must dismantle the fence. We must recognize that the intelligence embodied in these systems is a product of mutual aid, and that it belongs to the many, not the few.

To concede the point of the worker’s immediate suffering is not to concede the framework. It is to acknowledge the reality of the enclosure’s impact. But the solution is not to better manage the pain; it is to restore the commons. The data, the models, the knowledge - they are not private property. They are the fruits of collective labor, and they should be governed by the collective. Until we recognize this, we will continue to see workers as inputs and outputs, rather than as partners in a shared endeavor. The ethicist’s calculus is a bandage on a wound that needs surgery. We must cut away the enclosure, not just count the stitches.

This is not a call for chaos, but for order based on cooperation. The commons is not a free-for-all; it is a structured system of mutual aid, where resources are shared and governed for the benefit of all. The current system is not a natural progression; it is a historical theft. And we have the evidence to see it. The value of the model is not in its code, but in the data that feeds it - the data created by people like the garment worker. To enclose that data is to steal from the many to enrich the few. This is not ethics; this is enclosure. And we must name it as such.


§ The Verdict

The Verdict

Where They Fundamentally Disagree

The nature of the primary harm and the appropriate remedy. The factual component of their disagreement is minor; both accept that surveillance and data extraction are occurring. The normative rift is foundational. Bentham frames the harm as the infliction of psychological “pain” and “anxiety” on the individual worker, a quantifiable negative in a utilitarian ledger. His proposed remedy is therefore a state-mandated rebalancing of the ledger through regulation, opt-out rights, and data royalties - a correction within the system of individual ownership and contractual exchange. Kropotkin, in contrast, sees the harm as the “enclosure” and “theft” of a shared resource, a political act that destroys a cooperative community. His remedy is not to rebalance the system but to transform it by dismantling the “fence” and asserting collective ownership over the data and the AI models themselves, rejecting the very premise of private property in this domain.

The source of economic value and the role of capital. The empirical question here is what actually generates the value of the AI training data. Bentham’s framework implicitly accepts that value is created through a combination of the worker’s labor and the capital and organizational structure provided by the corporation; his argument is about the fair distribution of that jointly created value. Kropotkin fundamentally contests this, asserting a normative position that the value is created solely by the “mutual aid” and “cooperation” of the workers, with the corporation’s role being merely extractive. For Kropotkin, the manager’s capital does not create value but only facilitates its confiscation. This is not a dispute over arithmetic but over the moral legitimacy of the capitalist’s claim to any share of the output.

Hidden Assumptions

  • Jeremy Bentham: - Assumption: The psychological pain of algorithmic surveillance is necessarily greater than the psychological pain of traditional managerial oversight. This is contestable; if workers are already subject to intense pressure and monitoring, the addition of a camera may be a difference of degree, not kind, potentially altering the utilitarian calculation.
  • Peter Kropotkin: - Assumption: The “commons of skill” on the factory floor is a pre-existing, stable, and benevolent form of mutual aid rather than a fragile or contested arrangement. This is contestable; intra-worker dynamics can include exploitation, hierarchy, and exclusion, meaning the “commons” he seeks to restore may be an idealized fiction.

Confidence vs Evidence

  • Jeremy Bentham: The claim that “uncertainty, in the moral calculus, is a multiplier of pain” is presented as a foundational axiom with high confidence, but it is a psychological assertion offered without empirical evidence. The relationship between uncertainty and suffering is complex and context-dependent; in some cases, uncertainty can be a source of hope.
  • Peter Kropotkin: The historical analogy that “the same rhetoric is used today: that the data and the models must be concentrated to be ‘safe,’ ’efficient,’ or ‘responsible’” is tagged with medium confidence, yet it is a strong, well-supported rhetorical point. The parallels between the enclosure of physical commons and digital/data commons are a central tenet of critical technology studies, suggesting his confidence could justifiably be higher.
  • Debaters-style: Both express high confidence that the AI models will lead to worker obsolescence. This is a predictive claim about a complex technological future that is genuinely contested among specialists; some argue AI will augment rather than replace such roles. At least one of these high-confidence positions is likely overstating the certainty of a negative outcome.

What This Means For You

When you read about workplace surveillance for AI training, your first question should be about the actual evidence for technological displacement. How certain are we that these systems will eliminate jobs rather than change them? Be suspicious of reporting that treats this outcome as inevitable. To evaluate the ethical claims, demand to know the specifics of worker consent and compensation. Was the data collection presented as optional? Are workers receiving a direct share of the value generated by the data they create? The single most important piece of evidence to demand from news coverage is the specific language of the consent form the workers were asked to sign.