India unveils landmark AI governance guidelines — know what changes for startups, developers – Crypto News

The Ministry of Electronics and Information Technology (MeitY) on Wednesday published a seven-point artificial intelligence (AI) governance framework under the IndiaAI Mission—the nodal body tasked with developing built-from-scratch AI models native to Indian languages.

In the framework, the IT ministry has detailed that corporates and startups developing AI models and applications in India will need to use trusted data sources and offer transparency in the model’s workings as their key features.

The framework also details that ‘people first’ AI projects will be the ideal development area for companies, reiterating the government’s stance of using AI applications and models for public services.

The Centre has so far approved 12 startups under the AI Mission, offering $112 million to IIT Bombay-backed entity BharatGen in September. The framework will be applicable to the chosen startups under the mission, as well as other entities building AI in India.

A senior government official with direct knowledge of the matter told Livemint that while the framework released is not binding or applicable legally, it may set the tone for the regulatory approach that India is taking with AI.

“This is a set of expert committee observations and recommendations that suggest that if we want to pursue regulating AI in future, this is the approach that we can take. It also underlines that we must not look to establish a new law for AI unless absolutely necessary, and I think in a way it validates the approach that the government has taken so far,” the official said.

A second official added that the framework also clearly shows that the focus for the Centre will remain on building public utilities, such as the India AI Mission, which entails the use of public money for incentivization of the sector.

“This is important, because the governance framework does not necessarily apply to commercial entities building out of India,“ Deepro Guha, associate director for public policy at think-tank The Quantum Hub, said, adding that the framework offers ample insight into how regulations may evolve in future for AI.

“The techno-legal approach that the framework underlines is an interesting take on how the sector may evolve in future, and how regulations will adapt to a future where India may develop digital public infrastructure for AI. The framework has also offered ample thought into how laws will be amended to accommodate the advent of AI—such as the use of copyright laws to deal with infringement issues. Otherwise, it is a non-binding directional document that may not necessarily dictate how the industry will evolve in India,” he said.

Core principles and objectives

The guidelines are designed to strike a balance between innovation and accountability, ensuring that AI drives inclusive growth and maintains public trust.

The framework is focused on seven key principles: Trust, People First, Innovation over Restraint, Fairness & Equity, Accountability, Understandable by Design, and Safety, Resilience and Sustainability.

Dhruv Garg, cofounder and partner at India Governance and Policy Project (Igap), added that the framework “reflects a mature understanding of how fast the technology is evolving”.

“It’s a light but deliberate framework, prioritizing coordination and standards over control. Some concepts within this document would definitely need to align with the first principles of AI around privacy, rights-based approach, and rational platform governance. The next step will depend on what Meity, as the nodal ministry, identifies as its priority areas—be it institutional capacity, safety testing, or standard-setting,” he pointed out.

“A lot more coordination and consultation will be needed across ministries, regulators, and industry and other stakeholders before India can move toward a standalone AI law, sectoral regulations, or integrate AI provisions into any future IT Act refresh. Building that consensus will be key to ensuring that regulation keeps pace with innovation,” Garg added.

Move to boost AI infrastructure

The document highlights significant steps the government has already taken to build the national AI infrastructure backbone.

Compute Power: Over 38,231 GPUs (Graphics Processing Units) are being made available to startups, researchers, and developers at subsidised rates. A secure GPU cluster housing 3,000 next-generation GPUs is also under construction for sovereign and strategic applications.

Data & Models (AIKosh): The AIKosh platform has onboarded 1,500 datasets and 217 AI models from 34 entities across 20 sectors. It provides permission-based access, allowing contributors to retain control over data usage while facilitating AI development.

Startup Support: Four startups are being supported in the first phase to develop India’s sovereign models. They will receive credits and funding covering up to 25% of compute costs, provided through a mix of grants (40%) and equity (60%).

Governance and regulatory structure

The framework details a robust structure integrating governance, regulation, and oversight through the establishment of key bodies:

• AI Governance Group (AIGG): Responsible for overall governance.
• Technology and Policy Expert Committee (TPEC): Provides technical and policy support.
• AI Safety Institute (AISI): Focused on safety and risk mitigation.

The central government’s Principal Scientific Adviser Ajay Kumar Sood emphasised the core ethos of the framework, stating that the guiding principle of India’s AI approach is “Do No Harm.”

He noted that India’s framework will combine innovation sandboxes and risk mitigation mechanisms within a flexible and adaptive regulatory environment, ensuring that AI innovation thrives without jeopardising societal safety.

MeitY Secretary S Krishnan added that these guidelines mark a significant step towards human-centric and responsible AI development. He highlighted that India’s approach, anchored in its Digital Public Infrastructure, will position the country as a global example in safe AI adoption.