India Releases National AI Governance Guidelines
India's government has released new AI Governance Guidelines designed to balance innovation with regulation. The framework leverages existing IT and data protection laws, introduces a system of graded liability for AI-related incidents, and establishes a national AI Safety Institute and Governance Group. This move represents a significant step in formalizing India's national approach to AI oversight and sovereignty.
- The framework is anchored in seven guiding principles, or "Sutras," including "Innovation over Restraint," "People First," and "Accountability," which were first developed for the financial sector by a Reserve Bank of India committee. - A key aspect of the strategy is the pursuit of "sovereign AI," which aims to reduce reliance on foreign technology through initiatives like the IndiaAI Mission, a program with an outlay of over $1 billion. - The guidelines were unveiled by the Ministry of Electronics and Information Technology (MeitY) and formally launched by Prof. Ajay Kumar Sood, the Principal Scientific Adviser to the Government of India. - This approach avoids creating a new omnibus AI law, instead opting to adapt existing legal frameworks like the IT Act of 2000 and the Digital Personal Data Protection Act of 2023. - The policy establishes a new institutional architecture, including a multi-ministerial AI Governance Group (AIGG) for policy coherence and a specialized AI Safety Institute (AISI) for technical validation and risk assessment. - India's model is positioned as a "third path" in global AI governance, distinct from the US private-sector-led approach and China's state-directed model, emphasizing scalable applications built on digital public infrastructure. - The release of the guidelines was timed ahead of the AI Impact Summit in New Delhi, an event designed to be the first major global AI summit hosted in the Global South. - To support domestic innovation, the government's IndiaAI Mission is building a national compute infrastructure with a goal of 10,000+ GPUs and a data repository called AIKosh, which already hosts over 9,500 datasets.
