In the global race for artificial intelligence supremacy, data is the new oil. Now, India is charting its own course with a new handbook designed to unlock vast datasets for AI development, choosing agile guidelines over rigid new laws.
The Great Data Dilemma
Artificial intelligence
models are only as smart as the data they are trained on. To build powerful, unbiased, and effective AI, developers need access to massive and diverse datasets. However, sharing this data creates significant challenges. Institutions like hospitals, banks, and government agencies hold valuable information but are rightfully concerned about privacy, commercial confidentiality, and losing control over their assets. This friction—the need for data versus the need for security—often leads to a 'data deadlock', slowing down innovation. India's new 'Handbook on Data Empowerment and Protection Architecture (DEPA) for AI Training' aims to solve this very problem.
Inside the DEPA Handbook
Released by the Ministry of Electronics and Information Technology (MeitY) in collaboration with the iSPIRT Foundation, the handbook introduces a techno-legal framework rather than a new law. It proposes a system with three key players: 'Training Data Providers' (like banks and hospitals), 'Training Data Consumers' (AI developers and researchers), and 'Data Collabs', which will act as commercial intermediaries to pool data for specific sectors. A core principle of the framework is a paradigm shift in how data is accessed. As Principal Scientific Adviser Ajay Kumar Sood explained, the idea is that 'the model travels to the data, rather than the data travelling to the model'. This means AI models can be trained in secure, controlled environments without the raw data ever leaving the provider's possession, significantly enhancing security.
A 'Soft Law' Strategy for Innovation
India's decision to issue guidelines instead of enacting a new, overarching AI law is a deliberate strategic choice. This 'lightweight' approach is designed to be agile and adaptive, encouraging innovation by avoiding the potentially restrictive and slow-moving nature of formal legislation. The government has indicated that existing laws—covering information technology, data protection, and consumer rights—can be used to manage many AI-related risks for now. This positions India differently from regions like the European Union, which has pursued comprehensive, legally-binding regulations like the AI Act. The handbook encourages voluntary adoption of ethical codes and best practices, signalling a trust in industry-led self-regulation to build a responsible AI ecosystem.
Beyond Individual Consent
One of the most discussed aspects of the handbook is its approach to consent. For building large-scale AI models, the framework suggests that seeking individual consent for every piece of data is impractical. It distinguishes between AI 'inference' (using a model on one person's data, which requires consent) and AI 'training' (building a model from aggregated data). To protect individuals without their explicit consent for training, the handbook relies on technical safeguards like de-identification and differential privacy, a technique that adds statistical 'noise' to data to prevent any single person's information from being identified. This approach moves the responsibility of protection from individual consent to robust institutional and technological safeguards.
What It Means for Business and the Road Ahead
For Indian startups and corporations, this handbook provides a much-needed blueprint for data collaboration, potentially leveling the playing field against global tech giants with massive data stores. By establishing common protocols, it can reduce uncertainty and accelerate research and development in critical sectors like healthcare and finance. However, the voluntary nature of the framework raises questions about enforcement and accountability. Success will depend on widespread adoption and the effectiveness of the proposed self-regulatory organizations. The government has maintained that it will not hesitate to legislate if the need arises, making this handbook a foundational—but potentially first—step in India's evolving AI governance strategy.
