The Unseen Wages of India's AI Revolution
In Tamil Nadu, men and women who earn their living assembling electronics parts are being watched in ways that go beyond ordinary productivity monitoring. Motion sensors on factory floors log their movements. Their hand gestures, the speed at which they slot components onto circuit boards, the micro-adjustments their bodies make to accommodate defective parts — all of it is being fed into datasets used to train industrial robots. The workers are not asked. They are not compensated beyond their wages. The data belongs to someone else.
A report published this week by Scroll.in documented how major technology companies — Apple, Google, and others whose supply chains run through Indian manufacturing hubs — have quietly incorporated Indian worker data into the datasets used to automate factory tasks that humans currently perform. The report drew on interviews with workers, labour activists, and former employees of the third-party firms that operate the assembly lines on behalf of the brands. What it described was not illegal. That is precisely the problem.
The same week, the Central Bureau of Investigation widened its inquiry into leaks of the National Eligibility-cum-Entrance Test, India's standardised medical college entrance examination. According to the Indian Express, the CBI had questioned a doctor and the son of the founder of a coaching centre at the centre of the alleged racket. The probe had expanded to include searches for a middleman described as a key figure in facilitating the leak. The examination, taken by nearly two million students each year and widely understood to be the single most consequential assessment in a young Indian's life, has been compromised. The specific mechanics remain under investigation, but the psychological damage to millions of students and families who trusted the process is already done.
These two events have no obvious connection in the headlines. But they share a structure: in both cases, a system that Indians were led to believe served their interests has been revealed to be operating primarily for someone else's gain. The factory floor extracts body-movement data from workers who have no contractual claim over it. The examination system extracts trust from students who have no recourse when it fails them. Both are forms of dispossession — one material, one existential — and both are happening at a moment when India's political class is largely silent on the terms of the bargain the country has struck with global technology capital.
The Data That Was Never Yours
The practice Scroll.in chronicled — using human demonstration data to train robotic systems — is not unique to India. Factories in Vietnam, Mexico, and China serve similar functions for the same companies. What makes the Indian case notable is the scale of the workforce involved and the opacity of the arrangements. The third-party labour firms that employ the factory workers typically operate under agreements that vest intellectual property rights in whatever the factory produces — including data generated during production — to the brand or its nominated contractors. Workers told Scroll.in they had no idea their movements were being recorded for this purpose. Many did not know what a neural network was.
Labour rights advocates have a name for this dynamic: data colonialism. The term describes the extraction of value from populations who do not share in the ownership of the resource being extracted. It is a useful frame, though the more precise description is simpler and more uncomfortable: these companies needed real human dexterity to train their machines, they located that human dexterity in places where labour was cheap and regulatory oversight was weak, and they structured their contracts to ensure that the data generated belonged to them. No conspiracy is required. The incentives are sufficient.
There is a counterargument, and it deserves engagement. Proponents of this model note that automation ultimately displaces dangerous or repetitive work, and that the jobs being trained away from humans today would have been automated anyway through simulation-based training. They argue that the workers on these factory floors have employment precisely because the work has not yet been automated — that the data collection extends their employability rather than ending it. These are not trivial points. The automation of assembly work is a decades-long structural trend, not a policy choice that can be reversed by altering data-sharing agreements on the shop floor.
But they do not answer the core question: who captures the value when a worker's physical intelligence is translated into machine code? In the current arrangement, the answer is a shareholder in Cupertino or Mountain View. The worker receives a wage. The data goes to the model. The model's output competes with the worker's job.
When the Examination Fails
The NEET scandal operates on a different axis but arrives at a structurally similar place: a system designed to allocate opportunity has been captured by actors who treat it as a resource to be exploited. The leak, if confirmed on the scale some initial accounts suggested, would not be an isolated failure. It would be evidence that the examination has become an object of extraction — a thing from which rent can be taken — rather than a neutral instrument of assessment.
The Indian Express reported that the CBI investigation had widened to include a doctor and the son of the founder of a coaching centre in Rajasthan, and that authorities were searching for a middleman who allegedly facilitated the leak. What the sources do not yet establish is the scope of the compromise: whether a handful of candidates received advance access to questions, or whether the leak was systematic and widespread enough to invalidate the results for a significant cohort. That ambiguity matters enormously for the students involved, and it is the kind of ambiguity that Indian examination authorities have historically been reluctant to resolve quickly or transparently.
The trust damage is not easily repaired. For a student from a rural district who has prepared for NEET across two years of intensive coaching — often at considerable financial cost to their family — the examination represents more than a test. It is a claim on a future that the state has implicitly promised to allocate meritocratically. When that promise is broken, the damage is not just to the individual student but to the legitimacy of the entire institutional framework that connects educational attainment to social mobility.
The Silence at the Top
What is missing from both of these stories is not evidence — the evidence is available, in the Scroll.in reporting and the Indian Express coverage — but political attention. Neither the Narendra Modi government nor the principal opposition Indian National Congress has made the terms of India's technology-labour bargain a subject of genuine policy debate. Data rights for workers in export-processing zones are not on the legislative agenda. The governance of national examinations has not been the subject of structural reform proposals from any major party.
This is not a coincidence. India is competing actively for technology investment, and governments in New Delhi and the state capitals are aware that the labour cost advantage India enjoys depends partly on regulatory frameworks that give global buyers more control over the production process than comparable arrangements in higher-cost jurisdictions. Asking whether that control should include the unlimited use of worker-generated motion data is a question the investment pitch does not invite.
The political economy of silence is not unique to India. Comparable dynamics play out in Vietnam, Indonesia, and Bangladesh, where similar factory-floor data extraction occurs without local political consequence. But India is a democracy with a large and increasingly vocal labour movement, a free press, and an active judiciary. That these institutions have not yet generated meaningful pressure on the terms of data extraction from manufacturing workers is a measure of how technical the question appears — and how deliberately the companies involved have kept it technical.
The examination scandal has generated more political noise, partly because it is easier to understand and partly because it affects middle-class families who vote. The data extraction question is harder to dramatise: no individual worker has a visible grievance, no single piece of evidence proves harm, and the harm — to the extent it materialises — will arrive gradually as automation reduces the number of factory jobs available to the next cohort of workers.
A Country on the Receiving End
India's position in the global technology order is changing in ways that are not fully captured by the celebratory narratives about its software industry and its digital public infrastructure. The country has indeed built remarkable systems — the UPI payment infrastructure, the Aadhaar identity platform — that demonstrate genuine institutional capacity. But the other face of that capacity is that India has become a preferred site for the unglamorous labour that the technology revolution requires: the assembly work, the content moderation, the data annotation, and now, as Scroll.in reported, the physical demonstration data that trains the next generation of industrial automation.
There is nothing inevitable about this arrangement. Countries that have negotiated more assertively with technology investors — including some in Southeast Asia — have secured better terms on data ownership, local processing requirements, and technology transfer. The infrastructure to demand those terms exists. What is lacking is the political will to use it.
The workers on those factory floors, and the students who sat the NEET examination last week, are not abstractions. They are the population that Indian development policy is supposed to serve. When the systems meant to empower them are instead mechanisms for extracting value from them, the deficit is not merely economic. It is a failure of imagination at the top of a political class that has decided the terms of India's engagement with global capital are not open for negotiation.
They are, for now, correct. But that is a choice, not a constraint — and it is a choice that deserves far more scrutiny than it is currently receiving.