Office of Principal Scientific Adviser Government of India

AI governance is likely to evolve through a collaborative and interconnected global ecosystem rather than a single centralized regulatory authority. This evolving landscape will include national regulatory frameworks, international standards bodies, safety research institutions, digital infrastructure platforms, and voluntary industry commitments working together to support responsible AI development. In this context, the goal should be to promote interoperability and cooperation, enabling governance frameworks to function effectively across borders while allowing countries the flexibility to adapt them to their own economic priorities, social contexts, and developmental needs. The Global Dialogue on AI governance is a forward and welcome step towards achieving consensus for effective AI governance mechanism.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

Governments around the world are adopting diverse approaches to AI governance. Some jurisdictions emphasize rights-based regulation and product safety standards, while others rely on decentralized regulatory frameworks, sectoral oversight, or stronger state-led controls. At the same time, many countries are experimenting with hybrid approaches that combine regulatory safeguards with innovation incentives, public investment. There is urgent requirement to come together and discuss on minimum parameters required to ensure that AI deployments are safe trusted and secure , the cross border deployments take social ethical and cultural nuances in to account and interoperability is excercised for making AI inclusive for all.

There is requirement to discuss about AI induced risks, risk taxonomy, and dynamic and living standards to ensure that AI , across its life cycle, remain safe and secure for all. Old pattern of static technological standards will no longer be valid and new mechanisms of minimum parameter assurance based living and agile standards have to be dveloped. There is requirement for discussing standards at length.

There is requirement of making AI accessible to all in a safe , secure and inclusive manner. India's approach to AI governance aims to promote innovation while placing the principle of "Do Not Harm" at its core. India has proposed a techno-legal framework for AI governance that integrates legal instruments, regulatory oversight, technical safeguards, and digital infrastructure. The underlying principle is that since the AI systems operate at machine speed, governance mechanisms must also function through embedded technical controls rather than relying exclusively on ex-post regulatory enforcement. Our experiences with population scale DPI deployment indicate that effective oversight of AI systems can be achieved by combining technological mechanisms with legal and regulatory frameworks. We are open to discuss this further in the larger interest of the group.

International cooperation should need to focus on developing shared norms, auditable parameters, measurable safeguards and interoperable standards, that can support responsible AI development and deployment across boundaries while remaining adaptable to different regulatory environments. The AI dialogue can play an important role towards achieving this goal.

Existing mechanisms like ITU's AI for Good network, UNESCO's network of AI ethics experts without borders, the UN AI advisory body , various working groups under ISO, ITU, OECD etc could be leveraged for having systematic progress.

Thematic working groups with experts from acadeima , industry , civil society and other could be created. Virtual and focussed discussions within and across working groups will help in creating base working documents. These couments could further be discussed at national levels and finally by multilateral expert group.

Vulnerable groups like children, women , old age people and people from marginalised communities from developed and under developed nations are not appropriately represented in such dialogue. People from diffrent intersections have varied requirements, ofthen these are overlooked. In case of AI such under representation may further enhance existing disparities and gaps which may futher lead to extended biases and inequalities.

The thematic working documents prepared in a manner ,I have suggested at point 15 above, could be made open for consultations at all levels . Submission of comments on above documents could be called for in a time bound manner . Technology (Big data and AI) could be used for sorting very large ammount of comments received. Final outcome document could be vetted by multilateral expert group.

India is aimig to adopt a unique techno-legal framework for AI governance. A key component of India's approach is lifecycle-based AI governance, which recognizes that risks can arise at every stage of an AI system's development and deployment. Effective governance therefore requires oversight from the earliest stages of data collection through model development, deployment, and inference. In addition, as agentic AI systems gain greater autonomy, additional controls such as agent identity frameworks, authorization protocols, behavioural logging, and kill-switch mechanisms, etc. will be needed. Therefore it is important that collectively such technology development to be promoted which can be built within the AI systems to ensure that complete AI value chain is secure by design. Besides, having liability framework and human oversight need to be an important aspect of AI governance.