The human firewall against AI exploitation

Across India, thousands of workers are refusing to turn their labour into free training material for AI models. Their fight is about protecting their rights

Artificial intelligence is now omnipresent. In every walk of life, AI has made an intervention. Though it has several advantages, it also has a downside. It may well replace humans in various industries. And the irony is that it needs the help of those very workers it would replace. Every AI model must be trained first, and this has to be done by those who are presently doing that work. Every sophisticated AI model learns from billions of pieces of human-created content and from the painstaking work of countless annotators, translators, coders, writers and reviewers. Without this invisible workforce, AI simply cannot improve.It is therefore significant that many Indian workers are beginning to resist attempts to use their labour to train AI systems that may ultimately replace them. The concern is understandable. A content moderator who labels harmful posts, a language expert who translates documents, or a software engineer who corrects AI-generated code is not merely completing a task. They are transferring valuable expertise to machines. Once an AI system masters that knowledge, the demand for the same human skills may gradually diminish. Workers are effectively helping to build their own competitors. India occupies a unique position in this debate. It has become one of the world’s largest suppliers of AI training labour, employing hundreds of thousands in data annotation, linguistic validation, customer support, quality assurance and digital content moderation. Yet most of these workers have little say over how their contributions are used.

Knowledge is a form of intellectual capital. If companies profit enormously from AI systems built upon human expertise, there is a compelling argument that the individuals supplying that expertise deserve transparency, fair compensation and meaningful consent. Otherwise, AI risks becoming another example of technological progress that concentrates wealth while socialising its costs.

Can such practices be legally stopped? The answer is complex. Existing Indian labour laws do not specifically regulate AI training. Copyright law protects original creative works but not necessarily the knowledge or judgement workers apply during routine employment. Contract law generally favours employers where employees have already agreed to broad usage rights. However, India’s Digital Personal Data Protection framework and evolving discussions around AI governance could provide opportunities to strengthen worker protections. First, India needs clear legislation requiring companies to disclose when employees’ work is being used to train AI models.

Transparency should become a legal obligation rather than a corporate choice. Second, workers should have the right to informed consent wherever feasible, especially when their specialised expertise is being incorporated into commercial AI systems beyond the scope of their original employment. Third, labour laws should be updated to recognise AI-assisted workplaces. Finally, India should establish an independent AI ethics and labour commission to balance innovation with workers’ rights. Technological progress should never come at the cost of human dignity.