Research

Governance · 19 August 2026

AI-Enabled Digital Public Infrastructure

Independent working paper: six operational controls and a control loop so AI sitting on population-scale DPI remains visible, interruptible, and accountable.

A long empty civic corridor of pale stone with unlabelled service windows

This is an independent technical working paper (Version 1.0, August 2026). It is not an official publication of the United Nations, UNDP, or the DPI Safeguards Initiative. Digital Public Infrastructure is becoming how societies establish identity, make payments, exchange data, and deliver services at population scale. AI is moving from standalone applications into those same systems. The question is how AI can sit on DPI without weakening safety, inclusion, agency, accountability, and trust.

The central proposition: AI governance should become an operational component of DPI safeguards wherever AI can materially influence access, decisions, services, data, security, economic participation, or public trust. A model error in an isolated enterprise application may affect a department. A model error integrated into foundational DPI may affect an entire population. Governance intensity should increase with scale, autonomy, interconnectedness, and consequence.

Six layers: (1) Purpose and public value, do not introduce AI merely because it is available. (2) Risk and impact assessment, individual, institutional, and societal, including reversibility. (3) Human oversight architecture, awareness, authority, capacity, and intervention; a person merely appearing in a workflow is not sufficient. (4) Technical and procurement controls, architecture, data, security, auditability, supplier obligations. (5) Accountability and redress, decision rights and accessible challenge. (6) Continuous monitoring, measurement, and learning.

The AI-DPI Control Loop is Identify → Assess → Design → Test → Approve → Deploy → Monitor → Intervene → Learn. It should repeat whenever model, data, use-case, risk, or context changes. A theoretical override that cannot be executed under operating conditions is not a meaningful safeguard. Algorithms do not replace institutional accountability.

Procurement is an early chance to make safeguards contractual rather than aspirational. An inclusion test should examine language, literacy, disability, age, connectivity, affordability, and contestability. For consequential decisions, people should understand whether AI is involved, what role it plays, who is accountable, and how an outcome can be challenged.

The ultimate test of responsible AI is not how intelligent the system becomes. It is whether people remain protected, informed, empowered, and able to seek meaningful human accountability when the system affects their lives.

First published on n-consultancy.blogspot.com. This article was first published on my research blog and is updated here.