Unseen labour, exploitation: the hidden human cost of Artificial Intelligence
Context
Artificial Intelligence (AI) appears highly automated and efficient, but its accuracy relies heavily on the invisible labour of poorly paid workers in developing countries. From data annotation to content moderation, these workers — often called “ghost workers” — face exploitation, poor working conditions, and health risks, while their contribution remains unrecognised.
Invisible Human Backbone of AI
- AI and ML systems cannot interpret raw data without human input.
- Data annotators label text, images, audio, and video to train models.
- Example:
- LLMs like ChatGPT/Gemini rely on annotators for supervised learning & reinforcement learning.
- Self-driving cars depend on labelled video data to distinguish objects.
- The higher the quality of data, the more human labour is required.
Outsourcing to the Global South
- Major tech firms outsource annotation to workers in Kenya, India, Pakistan, China, Philippines.
- Workers face:
- Low wages (sometimes < $2/hour).
- Long hours and strict deadlines.
- No recognition of their contribution.
- Often hired through intermediary gig platforms, making labour fragmented, insecure, and opaque.
Types of Data Labelling
- Basic tasks: no expertise needed.
- Technical/niche tasks: require subject expertise.
- Issue: Non-experts are often forced to label medical or technical data → leads to errors in AI outputs.
Human Cost of ‘Automation’
- Content moderation: Workers manually review graphic, violent, pornographic material → leads to PTSD, anxiety, depression.
- AI-generated audio/video: Voice actors, performers (including children) supply the base data.
- Workers face surveillance, job insecurity, and union suppression when raising concerns.
Exploitation and Worker Testimonies
- 2024: Kenyan workers wrote to U.S. President Joe Biden describing AI labour as “modern-day slavery.”
- Complaints included exposure to extreme content, violations of labour laws, dismantling of unions, and systemic exploitation.
- Lack of transparency: workers often don’t even know which tech company they are serving.
Systemic Issues
- Fragmented gig economy → pay per microtask, no long-term contracts.
- Excessive surveillance and constant threat of dismissal.
- AI’s rapid progress is built on exploited, invisible labour chains.
Conclusion & Way Forward
- Advancement of AI is powered by “ghost workers.”
- Their invisibilisation benefits tech companies while perpetuating exploitation.
- Policy Recommendations:
- Stricter global and national regulations on AI labour supply chains.
- Transparency on outsourcing and subcontracting.
- Fair pay and recognition for workers.
- Stronger labour rights protections in the digital gig economy.
Related Posts
About the Great Indian Bustard
Posted on
PIB Analysis 28-11-22
Posted on
