The India Internet Governance Forum, the national chapter of the United Nations Internet Governance Forum, recently organised its fifth edition in India, with the theme “Advancing Internet Governance for an Inclusive and Sustainable Viksit Bharat”. The forum deliberated on India’s inclusive digital future, the role of digital infrastructure for resilient and sustainable growth, and the impact of artificial intelligence (AI). It addressed topics such as rural connectivity, open digital systems, domain name and DNS security, cybersecurity readiness, the evolution of India’s digital public infrastructure, data protection, content moderation, ethical deployment of AI and opportunities for young innovators. Key takeaways from the discussions…
Transforming AI for the planet and people: From frameworks to impact
AI governance today must bridge policy thinking and real-world implementation. While global and national frameworks are being developed, the key challenge lies in translating them into practical actions by governments, companies and users.
India’s AI governance guidelines
The India AI Governance Guidelines is a document that is intentionally framed as a governance framework rather than a regulatory one. This distinction is very critical: governance, as articulated by the committee, encompasses not only regulation but also adoption, diffusion, institutional capacity building and innovation enablement.
The guidelines have three defining characteristics. One, governance is not limited to compliance and restrictions. It also includes building ecosystems, enabling responsible adoption, supporting innovation and developing institutional capabilities across states, industry and society. Second, while informed by global developments in the European Union, the US and elsewhere, the guidelines begin from India’s own needs and aspirations, such as economic development, social inclusion, resilience, cultural representation and global competitiveness. This includes attention to local data sets, linguistic and cultural diversity, and risks such as caste and social bias that are specific to the Indian context. Third, the framework emphasises flexibility, adaptability and innovation, reflecting the belief that overly rigid rules could stifle beneficial uses of AI in a rapidly evolving technological landscape.
There are three practical imperatives for operationalising responsible AI:
- Prioritise: Companies operating in India must meaningfully prioritise Indian contexts, risks and laws, rather than treating India as a secondary market.
- Implement: Responsible AI principles must translate into concrete practices – red teaming, risk mitigation and sector-specific safeguards.
- Demonstrate: Crucially, companies must be able to demonstrate compliance and responsibility to regulators through evidence, audits and documentation.
The industry is calling for the creation of an AI incident reporting database, which would generate empirical evidence on risks and failures in the Indian context, and the establishment of an institutional ecosystem comprising an AI governance group, a technology and policy expert committee and an AI safety institute.
Industry transparency and foundation models
Transparency and accountability are not merely regulatory obligations but are in the industry’s own interest, as trust is essential for adoption. The existing industry practices include developer documentation and use guides, detailing model architecture, limitations and intended use cases; and model cards, now an industry-wide norm, which summarise model capabilities, risks, appropriate uses and increasingly, environmental impacts such as energy consumption and carbon emissions.
However, significant challenges remain. General-purpose models are designed to be deployed across diverse sectors such as healthcare, finance and education, each with different risk profiles.
Full disclosure of every potential use and harm is neither practical nor necessarily useful. Transparency standards must increasingly become contextual and sector-specific, rather than one-size-fits-all. Further, multi-stakeholder consultation is important to refine best practices as technology evolves.
Scaling AI: From pilots to ecosystems
There exists a gap between successful pilots and scalable solutions. Drawing on experiences in agriculture, healthcare and education, such as open agri-networks and health-focused AI models, it has been noted that India has reached an inflection point where foundational digital infrastructure (digital identity, payments and connectivity) is largely in place. Yet, scaling remains difficult due to several factors:
- Local capability gaps: While global technology firms can support pilots, long-term success depends on local governments, institutions and start-ups with the skills, champions and operational capacity to sustain and scale solutions.
- Procurement and institutional friction: Government procurement processes, procedures and adoption models can slow down or block scaling, even when pilots are successful.
- Compute and infrastructure constraints: Large-scale, real-time AI, especially voice-based systems for hundreds of millions of users, requires inferencing infrastructure that India does not yet fully possess.
There is a growing need for vendor-agnostic ecosystems, open standards and capacity-building initiatives such as start-up accelerators and compute-sharing programmes. Transparency initiatives such as content provenance tools (for instance, watermarking and verification systems) and industry collaborations on standards are positive steps forward.
India’s progress and gaps: A government perspective
India’s progress in leveraging AI for productivity and efficiency gains has been significant. Examples include summarisation, decision support and data analysis – tasks that dramatically reduce time and cost for individuals and organisations. It is important that similar gains are democratised across sectors:
- Agriculture: Empowering farmers with real-time, localised knowledge.
- Micro, small and medium enterprises (MSMEs) and retail: Enabling data-driven decision-making for inventory, sourcing and sales.
- Education: Providing AI tutors to address shortages of quality teachers and personalise learning in local languages.
- Healthcare: Extending diagnostic and advisory capabilities to underserved areas.
However, the benefits must be balanced against risks. Despite claims of responsibility, AI systems continue to exhibit serious failures, such as bias, harmful outputs and unsafe behaviour, stemming from incomplete data, cultural blind spots and opaque training processes.
The analogy of traffic regulation can be used: just as society accepts speed limits and safety norms for vehicles, AI systems operating at a population scale require enforceable guardrails, accountability and liability frameworks.
Unfinished agenda: Compute, sovereignty and skills
India’s efforts so far, while significant, can be seen as only scratching the surface. Key gaps include:
- Inferencing at scale: Voice-based AI services for hundreds of millions require massive, low-latency compute infrastructure.
- Indigenous AI tools: Reliance on foreign coding assistants risks transferring India’s cognitive advantage to global platforms. Building domestic alternatives is seen as critical for technological sovereignty.
- Human capital: Beyond compute, sustained investment in skills, champions and institutional expertise is essential.
Societal transformation, labour and power
Artificial Intelligence must be seen not merely as a tool but as a vector of societal transformation. Early signs that AI is disproportionately affecting entry-level workers and restructuring labour markets are already being seen.
Key risks include task-level automation leading to job displacement, expansion of precarious data and platform work, and concentration of compute, data and value in a few transnational corporations.
Governance must move beyond managing negative externalities and instead proactively shape outcomes. Key recommendations include:
- Labour market foresight: Establish observatories to track AI-induced disruptions sector by sector.
- Worker-centric regulation: Require AI impact assessments on labour and mandate worker consultations.
- Domestic innovation support: Public compute, open data sets and safeguards for MSMEs.
- Social protection and care policies: Recognising care as foundational to inclusive participation, especially for women.
- Democratic value distribution: Supporting cooperative and worker-governed AI platforms.
- Robust data governance: Mandatory data-sharing frameworks for public value creation.
Conclusion
AI governance is no longer about abstract principles alone. It is about institutions, infrastructure, labour, power and trust. India’s approach – balancing innovation with guardrails and national priorities with global alignment – can be seen as both an opportunity and a responsibility.
The transition from frameworks to impact will depend on sustained collaboration between the government, industry, academia and civil society, with a clear-eyed recognition that AI is reshaping not just markets, but society itself.
