Indian data workers train robots as AI firms seek real-world labour
A new photo-led report on Indian workers training AI robots points to a wider shift in the automation economy: the next generation of machines needs human labour before it can replace or reorganise it. The lead report documents workers in India generating training data for AI-enabled robots. Separate reporting on AI data work in India shows that annotators and moderators already perform low-visibility tasks for global technology supply chains, while Business Insider's reporting on Instawork shows robotics companies paying gig workers to record, label and support real-world tasks. The International Labour Organization's 2023 working paper argues that generative AI is more likely to augment many occupations than fully automate them, but it also stresses job-quality risks. For Europe, the European Commission's AI Act framework matters because AI used in employment and worker management can fall under high-risk rules, while general-purpose model providers face transparency obligations.
Verified by Validiris·📚 6 sources·🧠 AI-checked·🇧🇪 Belgian: MediumWhy you can trust this
About this story
India (the world's most populous country and a major IT-services and business-process outsourcing hub) is central to global AI data work. Al Jazeera (Qatar-based international broadcaster founded in 1996) published the visual lead on workers training robots. Instawork (San Francisco gig-work platform founded in 2015) is a useful comparator because it has built robotics data services around hourly workers. Instacore (Instawork's wearable camera system announced in 2026) records work tasks for robotics training. The International Labour Organization (United Nations labour agency founded in 1919) researches work standards and AI's labour-market effects. The European Commission (EU executive headquartered in Brussels) oversees EU AI policy. The AI Act (EU Regulation 2024/1689, in force since 2024) sets risk-based rules for AI systems in the EU. Ding Wang, Shantanu Prabhat and Nithya Sambasivan (human-computer interaction researchers) studied Indian data annotation work in a 2022 paper.
How to read this story
The history
AI's hidden workforce predates the current robot boom. Amazon Mechanical Turk popularised online microwork in 2005, and image-labelling firms later supplied datasets for autonomous vehicles and computer vision. Wang, Prabhat and Sambasivan's 2022 study of Indian annotation work found that annotators' work practices were shaped by clients and managers above them, not by worker autonomy. The ILO's 2023 working paper then framed generative AI mainly as an augmentation shock, while warning that job quality and fair transitions would depend on policy choices. Robotics adds a physical layer: machines need data about movement, objects, workplaces and human improvisation.
The geopolitics
Physical AI is becoming another front in the competition between US, Chinese, Indian and European technology ecosystems. Countries and firms that control data, chips, robotics talent and workplace deployment channels will shape productivity gains. India's role is not only as a software market; it is also a labour and data infrastructure for systems that may later compete globally.
Why now
The lead is timely because AI developers are moving from text-based models toward robots that need real-world training data. The European regulatory clock also matters: the Commission says the AI Act is moving through phased application, with full applicability due on 2 August 2026 and later deadlines for some high-risk systems.
What to watch
Watch for vendor disclosures on how robotics training data is collected, whether workers can opt in meaningfully, and how EU buyers document compliance. In Europe, the next practical test is how national authorities and the European AI Office interpret AI Act duties for employment tools, robotics products and general-purpose model providers.
International angle
The story sits inside a cross-border AI supply chain: training data can be generated in India, sold through technology vendors, and embedded in tools later deployed in Europe. The European Commission's AI Act framework gives Brussels a regulatory role, especially where AI systems affect employment, safety, education or worker management inside the EU single market.
What this means for you
Belgian employers considering AI-enabled robots or workplace AI should ask suppliers where training data came from, whether workers consented, how personal data was handled, and whether the system falls into an EU high-risk category. Workers and unions should treat robotics pilots as labour-organisation decisions, not only technology purchases.
What happens next
The next signals are regulatory and commercial. The European Commission says the AI Act becomes fully applicable on 2 August 2026, with later timelines for some high-risk systems. Robotics firms are expected to keep seeking real-world movement and workplace data. Belgium-based buyers should expect more vendor claims about human oversight, training data provenance and compliance as AI-enabled products enter workplaces.
Potential consequences
If robotics data collection scales, Belgian and European firms could gain more capable automation tools for logistics, care, retail and manufacturing. The risk is a two-tier AI economy in which low-paid workers abroad generate the data while higher-income markets capture productivity gains. Poorly documented training data could also weaken safety and bias controls. Conversely, transparent procurement and EU compliance pressure could push vendors to document data sources, worker consent and human oversight more clearly.
Opposing perspectives
- AI labour researchers
Wang, Prabhat and Sambasivan argue that data annotation should not be treated as a neutral technical input. Their 2022 study frames Indian annotation as work shaped by organisational power, client demands and limited worker voice, making welfare, career pathways and participation central to AI quality.
- Robotics data platforms
Instawork's executives argue that robotics data demand can create new work rather than simply erase existing jobs. The company's account presents gig workers as trainers, technicians and field support staff who help robots learn from complex real workplaces before deployment.
- International Labour Organization
The ILO working paper argues that the dominant near-term effect of generative AI may be augmentation rather than full automation, but says policy should focus on job quality, social dialogue and fair transitions because the distribution of benefits is not automatic.
Timeline
- 2005·Amazon Mechanical Turk helped popularise online microwork for small digital tasks.
- 2022-03-21·Wang, Prabhat and Sambasivan submitted their study on Indian data annotation work.
- 2023-08-21·The International Labour Organization published its working paper on generative AI and jobs.
- 2024-08-01·The EU AI Act entered into force, according to the European Commission.
- 2026-06-11·Al Jazeera published the photo-led story on Indian workers training AI robots.
Glossary
- AI Act
- The EU's risk-based artificial-intelligence regulation, formally Regulation (EU) 2024/1689, setting obligations for providers and deployers of AI systems.
- High-risk AI system
- An AI system used in sensitive areas such as employment, education, critical infrastructure or some product-safety contexts, subject to stricter EU obligations.
- General-purpose AI model
- A model capable of performing a wide range of tasks and forming the basis for many downstream AI systems.
- Data annotation
- Human labelling, rating or structuring of images, video, text, audio or actions so machine-learning systems can learn from them.
How this story developed
2 reports on this subject — earliest first. You are reading the highlighted entry.
- Indian data workers train robots as AI firms seek real-world labour· You are here
- Chinese robot makers test AI cleaners in homes
Related to this story
Live connections from the Belgium Impulse ecosystem — not recommendations.
This briefing was prepared with AI assistance and reviewed by a Belgium Impulse editor before publication. methodology.