Indian Factory Workers Turned Unpaid Data Sources for AI Robotics

In garment factories on the outskirts of Delhi, supervisors have begun handing workers head-mounted cameras. Workers such as Lalita, a 32-year-old seamstress in Gurugram, are required to wear the devices for entire shifts. The cameras capture hand movements, machine interactions and even casual conversations, creating “egocentric” video that records work from the worker’s point of view.

The Guardian

Robotic systems need first-person footage of human labor to learn how to manipulate objects and navigate real-world environments. Industry analysts describe data as the biggest bottleneck for robotics, and India offers a unique combination of scale, diversity and labor density. The country’s large informal workforce and low-cost environment make it attractive for firms seeking millions of hours of footage.

The Guardian

• EgoLab – Indian data aggregator that supplies footage to Tesla.

The Guardian

• Tesla – CEO Elon Musk predicts 80 % of the company’s future value will stem from its humanoid robots.

The Guardian

• Humyn AI, FPV Labs, Micro1, Egodata, Neocambrian, XP Robotics, Objectways, Scale AI, CynLr – firms building data pipelines for robotics clients.

The Guardian

• Labellerr AI – founded by Puneet Jindal, collects and annotates egocentric video.

The Guardian

• Aapti Institute and Work Fair and Free Foundation – research NGOs monitoring the practice.

The Guardian

India accounts for roughly 35 % of the global data-annotation market, with about 60 % of revenues coming from U.S. clients. Companies estimate that training robust robots will require hundreds of millions to billions of hours of human activity footage. A U.S. firm paying $30 per hour for data collection can obtain comparable work in India for less than one-sixth of that cost, according to an anonymous technology founder.

The Guardian

Elon Musk has publicly stated that the bulk of Tesla’s future valuation will derive from its humanoid robots. Puneet Jindal emphasized India’s “workshop of the world” status, noting the unmatched combination of labor scale and diversity. An unnamed founder highlighted the cost advantage, while several data-collection firms argue that factories receive compensation for facilitating recordings, making additional worker payments unnecessary.

The Guardian

Researchers such as Geeta Thatra (Work Fair and Free Foundation) warn that consent is ambiguous when employment is insecure, and that workers may be unable to refuse cameras without fearing job loss. Privacy issues arise when cameras capture intimate moments, including bathroom visits. Madhumita Dutta (Ohio State University) stresses that workers generate “valuable digital assets” without awareness of downstream licensing or sale.

The Guardian

Lalita describes the cameras as “fun at first” but notes a shift to quieter conversations and heightened self-monitoring. She receives only a soft drink as a token, despite earning about $200 per month. In Bengaluru, mason Munazir earns $30-$40 weekly for recording his work, yet admits he “has no idea” how the footage is later used.

The Guardian

The Guardian’s investigation found no direct worker payments in six factories across five states, while companies claim factories are compensated. Conversely, informal workers like Munazir receive modest hourly wages from local contractors. No technology firm reported obtaining consent directly from workers; permissions were secured through factory management.

The Guardian

• “The way people mount a CCTV camera on a wall, they mounted one on us,” — Lalita, garment worker

The Guardian

• “South Asia remains the workshop of the world for many labour-intensive industries. If you’re trying to teach a robot how humans work, there are few places that offer the same combination of scale, diversity and density of human labour as India. On any given day, millions of workers are sewing garments, assembling products, sorting goods and performing tasks that robotics companies want machines to learn,” — Puneet Jindal, founder, Labellerr AI

The Guardian

• “double quotation mark A worker may appear to agree to wear a camera, but can they refuse without fearing consequences for their job?” — Geeta Thatra, researcher, Work Fair and Free Foundation

The Guardian

• “If they are unaware that their movements, skills and routines are being converted into datasets that can be licensed, sold or used to train commercial AI systems, they have little opportunity to negotiate compensation or object to downstream uses.” — Madhumita Dutta, Ohio State University researcher

The Guardian

• “We are not even getting our full worth for the work we do now,” — Lalita, garment worker

The Guardian

• “The data originates in a worker’s body and actions, but once extracted it no longer remains attached to them in the same way,” — Sarayu Natarajan, founder, Aapti Institute

The Guardian

Policymakers are urged to explore mechanisms such as royalties, profit-sharing or statutory data-ownership rights to ensure workers receive value beyond daily wages. Ongoing debates focus on establishing clear consent protocols, transparent data-use disclosures, and legal frameworks that treat egocentric recordings as labor assets rather than free inputs for AI development.

The Guardian