The article examines a growing practice in Indian factories where workers are asked to wear head-mounted cameras that record their daily activities. These first-person ("egocentric") recordings capture workers' hand movements, decision-making, and task execution, creating valuable datasets used to train AI models and humanoid robots. Many workers reportedly receive little or no explanation about how the footage will be used and are often unaware that their everyday skills are becoming commercial training data for global robotics companies.
India is emerging as a major hub for collecting this type of robotics data because of its large manufacturing workforce and comparatively low data collection costs. Companies specializing in AI data collection partner with factories to gather millions of hours of human activity, which are later cleaned, annotated, and supplied to robotics developers. The article notes that firms involved in this ecosystem work with international AI and robotics companies, reflecting the growing demand for real-world human demonstrations needed to train next-generation robots.
The report raises significant ethical and labor concerns. Workers interviewed expressed fears that they are effectively helping train machines that could eventually replace their own jobs, yet they receive only their normal wages and no additional compensation for generating valuable AI datasets. Labor advocates also question whether workers provide truly informed consent, given that declining to participate may not be a realistic option in many workplaces. Beyond compensation, concerns include workplace surveillance, privacy, ownership of behavioral data, and the potential use of recordings to monitor productivity or evaluate employee performance.
The article concludes that AI development is creating a new category of labor in which workers contribute not only physical work but also digital knowledge that can be monetized repeatedly through AI systems. Researchers and labor rights experts argue that existing labor laws are not designed for this emerging reality and suggest that governments and companies should explore stronger transparency requirements, explicit consent, and potentially new compensation models—such as royalties or profit-sharing—for workers whose expertise becomes part of commercially valuable AI training datasets. As robotics adoption accelerates, the debate is shifting from whether AI should be trained on human work to how workers should be recognized and protected in the AI economy.
