Across India, thousands of workers are being paid to record everyday activities using smartphones, GoPro cameras, and motion-capture devices attached to their heads. These recordings capture tasks such as slicing fruit, folding clothes, filling bottles, operating machinery, and other routine actions. The data is then used by AI companies to train robots and machine-learning systems to understand how humans interact with the physical world.
One example highlighted in the report is a young woman in Chennai who earns about ₹250 per hour by filming herself performing household chores. Similar projects are taking place in factories and specialized data-collection facilities, where workers generate what is known as “egocentric data” — first-person recordings that help robots learn human movements, hand coordination, and decision-making in real-world environments.
The work is creating a new source of income and supporting India's growing role in the global AI supply chain. Companies collecting this data argue that robots trained on human actions could eventually improve productivity, address labor shortages, and perform repetitive or physically demanding tasks. The demand for such training data is rising rapidly as technology firms race to develop increasingly capable humanoid robots.
At the same time, the trend has sparked debate about automation and employment. Many workers and observers worry that the knowledge being captured today could ultimately be used to automate the very jobs that provide livelihoods for millions. Online discussions have reflected concerns that workers are effectively helping create systems that may one day replace them, while others argue that new roles in robot supervision, maintenance, and AI operations could emerge alongside automation.
