Trust as the foundation: Reframing AI governance in the public sector

As artificial intelligence begins to guide welfare targeting, regulatory oversight and service delivery in India, that question becomes even more pertinent. AI can help governments work faster and make more informed decisions but when people cannot see or understand how those decisions are made, the system can start to feel distant or opaque. The shift, therefore, is not just from ‘digital systems’ to ‘AI-enabled systems.’ It is from ‘AI for All’ to something more demanding which is ‘AI you can trust.’

India’s early articulation of ‘AI for All’ through NITI Aayog’s National Strategy for AI (NSAI) was aspirational and forward looking.¹ It placed artificial intelligence at the centre of inclusive growth across health, agriculture, education, smart cities and mobility. That framing was important as it indicated an ambition for adoption of AI with trust and governance being core to it.

The India AI Mission and the development of national AI governance guidelines have taken the policy level conversation to the architecture level. The underlying recognition is although simple but critical in the sense that it stresses trust to be designed in the systems rather than just focusing on efficiency gains.

India’s approach to governing AI is gradually taking shape around a set of practical principles. A key recognition is that ethics, transparency and accountability cannot be treated as optional features; they are fundamental to building AI systems that people can trust.² The articulation of responsible AI Sutras as enumerated in the India AI Governance Guidelines – trust, human-centric, innovation over restraint, fairness and equity, accountability, explainability, safety, reflects a desire to remain technology neutral while still being normatively clear.

Equally important is the need for a whole-of-government approach. AI systems often rely on shared data infrastructure and interoperable platforms that cut across departmental boundaries. Fragmented governance could lead to uneven safeguards and potential blind spots. A coordinated model, supported by central oversight and common standards are therefore essential for effective risk management.

One of the essential acknowledgements is that the technological components and capabilities are still evolving. It is yet to be understood completely in terms of its impact on large scale adoptions, environment effects and impact on human values. A roadmap which could be phased out across institutional building, standard setting and regulatory framework would ensure that there is a scope to adapt without stifling technological innovation.

The articulation of the MANAV vision by the Hon’ble Prime Minister of India at the India AI Impact Summit adds an important political dimension to this discussion.³ By framing AI around values such as morality, accountability, national sovereignty, accessibility and legitimacy, the emphasis shifts from what technology can do to what it means for people. That shift is significant, because public trust is rarely built through technical documents alone; it strengthens when political leadership repeatedly signals that dignity, inclusion and citizens’ rights remain central.

The real test now lies with ministries and state governments. Translating high-level principles into day-to-day administrative practice requires careful choices. These include procurement frameworks, model validation processes, audit mechanisms, human oversight structures, and accessible grievance redressal mechanisms. It also requires civil servants who are equipped comfortable to question, interpret, and challenge technological outputs. It demands clarity on who is accountable when automated outputs influence high-stakes outcomes.