India Unveils AI Governance Framework, Opts For ‘Light-Touch’ Oversight

The Indian government on Wednesday released its much-anticipated AI governance guidelines, outlining a principle-based, light-touch framework for safe and responsible adoption of artificial intelligence across sectors.

Calling it a "foundational reference for all stakeholders", MeitY Secretary S. Krishnan said the framework would help balance innovation with accountability while relying largely on existing laws for governance. "The government has made it clear that it will react reasonably, this is intended to be light-touch regulation," Krishnan said.

The framework lays out guiding principles or "Sutras" and an action plan spread across short-, medium-, and long-term goals, emphasizing inclusive growth, risk mitigation, and alignment with global standards.

Proposed AI Structure

• An AI Governance Group (AIGG) will oversee policy coordination,

• A Technology & Policy Expert Committee (TPEC) will offer strategic guidance,

• An AI Safety Institute (AISI) will handle risk assessment, standards, and global engagement, and

• Sectoral regulators will enforce domain-specific compliance.

The document also calls for developing India-specific risk classifications, launching an AI incident reporting system, and embedding human oversight in high-risk applications.

On the infrastructure front, the government aims to expand compute and data access, support indigenous foundation models, and integrate AI with Digital Public Infrastructure (DPI) to enhance governance.

By encouraging regulatory sandboxes, voluntary safeguards, and MSME adoption of AI, the framework positions India's approach as "agile, inclusive, and innovation-driven."

Commenting on the development, Probir Roy Chowdhury, Partner, JSA Advocates and Solicitors, said, "For businesses, this translates into a mandate for improved accountability and compliance by design. Companies should focus on building robust risk assessment frameworks and consider adopting the voluntary compliance measures - I believe these will soon become the baseline for safe and trusted AI deployment."