Companies and contractors are increasingly paying people in India to film themselves while they work and share the footage for use in training artificial intelligence systems, the BBC has reported. The videos — intended to show real-world human activities and environments — are stitched into datasets that help AI models learn to recognise actions, gestures and workplace scenarios.
How the arrangements operate
Participants are typically recruited through online platforms or intermediaries and are compensated for supplying footage and completing simple upload or annotation tasks. The recordings are intended to provide diverse, realistic examples of daily and occupational activities that are difficult for companies to obtain through staged shoots or synthetic data alone.
Benefits and incentives
For many workers, taking part offers a source of extra income and may pay more than traditional data-labelling gigs. For AI developers, the material helps improve model performance and robustness by exposing systems to a wider range of lighting conditions, body types, work tools and home and workplace environments than might otherwise be available.
Privacy, consent and labour concerns
Human-rights advocates, researchers and labour experts say the approach raises significant ethical and legal questions. Critics point to risks around informed consent, the potential for footage to be repurposed for surveillance, and inadequate protections against reidentification. There are also concerns about whether participants fully understand how their images will be used and whether the compensation reflects the downstream value of the data.
Industry and regulatory context
Some firms involved in collecting such footage say participants are paid and that measures to de-identify or limit use are in place, but independent oversight is limited and regulatory frameworks are still catching up with rapid advances in AI. The growing reliance on real-world visual data has intensified calls for clearer standards on transparency, fair pay and data safeguards.
As AI systems are deployed across more applications — from safety-critical automation to consumer-facing services — the debate over how training data is sourced and governed is likely to become a central policy and corporate responsibility issue. Analysts say resolving tensions between innovation, workers' rights and privacy will be crucial to building public trust in next-generation AI technologies.