

On the first anniversary of the Global Digital Compact (GDC), Digital Cooperation Day gathered governments, standards bodies, UN agencies, civil society, and industry to answer a simple question: have we moved from principles to practice?
Hosted by the UN’s Office for Digital and Emerging Technologies (ODET) during UNGA High-Level Week, the day’s sessions suggested a real shift is underway—from high-level vision to the architecture, tools, and coalitions needed to implement it.
By Nthanda Manduwi • New York City • 22 September 2025
The GDC was adopted on 22 Sept 2024 as an annex to the Pact for the Future, establishing shared goals on connectivity, data governance, AI, and digital rights. Over the past year it has matured from a political text into a policy work program: member states and partners are piloting new forums, standards pathways, and capacity-building mechanisms to make cooperation real.
Two elements of that emerging architecture stood out throughout the day:
The “Universal DPI Safeguards — Guardrails to Accelerate SDG Progress” session made the day’s most consequential point: DPI has become governance infrastructure. The Universal DPI Safeguards Framework reframes “digital rails” as public institutions that must be rights-respecting and accountable across their lifecycle—design, deployment, operation, and evolution. The companion push to translate principles to outcomes (assessment guides, onboarding pathways, resource hubs) is helping countries measure whether their DPI actually protects privacy, inclusion, and equity in practice.
Financing and capacity remain the hard constraints. That’s why the 50-in-5 effort—helping 50 countries design, launch, and scale core DPI components by 2028—surfaced repeatedly, including at the evening “DPI Cooperation in Motion” milestone event during High-Level Week. It is becoming the practical engine room for country-led DPI implementation with safeguards “baked in,” not bolted on.
Synthesis: The DPI conversation has matured: from “should we build national stacks?” to “how do we build them safely, inclusively, and interoperably—and how do we prove it?” The center of gravity is shifting from abstract guardrails to verifiable outcomes.
Panels on AI for inclusive economies and international standards underscored how far the governance debate has moved since 2024: less fixation on sci-fi risks, more on the mundane (and essential) work of standards alignment, testing and assurance, and capacity building—especially outside a handful of advanced ecosystems. The HLAB-AI’s blueprint is informing that pivot with proposals for a standards exchange, capacity network, and global policy dialogue to reduce fragmentation and build shared baselines.
Still, the distribution problem is unresolved: compute, data, and research investment remain concentrated. Without financing and capacity mechanisms (including the proposed Global AI Fund) and support to localize standards and risk practices, global governance can become global gatekeeping. The day’s discussions were candid about that risk; the solutions on offer—distributed scientific assessment, interoperable standards, and pooled capacity—are steps toward more equitable global AI governance.
“The Global Digital Compact One Year On,” “Towards an Inclusive Digital Economy,” and “Localizing the GDC—Mayors for Digital Cooperation” all emphasized a simple truth: implementation is local. Cities are where digital services meet residents; national strategies succeed or fail based on local capacity, trust, and relevance. The GDC’s value now depends on whether it empowers municipal and national actors—to adapt shared principles to local contexts, not merely import models. (This same logic is explicit in ODET’s DPI and GDC briefings.)
The OHCHR-anchored session, “Data with Dignity at its Core — Protecting Privacy and Equality,” grounded the technical agenda in human rights. OHCHR’s 2025 reporting cycle warns that data-driven systems can reinforce discrimination and exacerbate inequality unless privacy, non-discrimination, and due process are designed (and enforced) end-to-end. That human-rights lens is now bleeding into DPI and AI efforts: measurement, redress, and accountability are being treated as features, not compliance afterthoughts.
Synthesis: Trust is not a comms problem; it is an institutional design problem. Embedding rights, redress, and transparency into DPI and AI—and proving it with public evidence—is essential for legitimacy.
The day also looked outward: WSIS+20 (the 20-year review of the World Summit on the Information Society) and related fora (IGF, ITU’s AI for Good, standards bodies) are aligning with the GDC’s priorities. Expect 2025–2027 to be a consolidation phase: stock-takes, interoperability work, and mid-term GDC reviews to check whether commitments are translating into measurable outcomes.
At UNGA80, I argued that Africa must be a creator of AI, not just a consumer; that we must understand before we regulate; and that without energy and financing, our AI ambitions remain rhetoric. Those convictions reverberated through Digital Cooperation Day—but with a new nuance: the global governance scaffolding (from DPI safeguards to AI standards and scientific assessment) is finally catching up to that vision. The challenge is to ensure these mechanisms are truly distributed—so countries can co-create, not merely comply.
What changed in Year 1
What must change in Year 2
A year ago, the GDC gave us a North Star. Today, we are building the constellation around it: DPI safeguards to anchor rights and inclusion; AI governance mechanisms to align standards and capacity; and processes (WSIS+20 and others) to stitch the ecosystem together. The next phase is less about declarations and more about delivery—verifiable protections, interoperable systems, and shared capacity that widen the circle of those who can build and benefit.
If we succeed, we won’t just connect people to digital systems—we will co-create digital institutions worthy of public trust.
If you’d like to go deeper into my journey — from Malawi, through the United Nations to Microsoft, you can find it in my books: