AI for social inclusion (HLPF side event)

Session: HLPF 2026 side event, “AI for Social Inclusion: Opportunities, Risks and Policy Priorities,” convened to feed the AI chapter of UN DESA’s forthcoming World Social Report
Hosts: UN DESA and the Permanent Mission of the Philippines to the UN (in its capacity as 2026 ASEAN Chair)
Moderator: Navid Hanif (Assistant Secretary-General, UN DESA)
Opening: The Deputy Permanent Representative of the Philippines (Ariel Peñaranda)
Panel: Xiaolan Fu (Professor of Technology and International Development, University of Oxford, and member of the UN High-Level Advisory Board on Economic and Social Affairs), Vijay Modi (Columbia University) and a youth representative of the Global Forum of Communities Discriminated on Work and Descent

Inclusion is not automatic

The premise of the session, stated at the top and again at the close, was that AI does not include anyone by default. Left to the market it tends to concentrate benefits and encode the biases in its training data; inclusion has to be designed in, through policy, governance and targeted measures. UN DESA framed the event as a preview of the World Social Report’s coming chapter on inclusion in the age of AI.

Navid Hanif, UN DESA: “Without inclusive policies, AI can and will reinforce its biases. It will widen inequalities and deepen existing digital divides. The moat will grow if we do not work in building bridges to address these gaps.”

The opportunity and the divide

The panel did not doubt the upside. Xiaolan Fu of Oxford pointed to AI in drug discovery and diagnostics for underserved regions, and to tools that let young founders and researchers get their ideas evaluated cheaply; the Philippines cited AI in disaster modelling and remote-island public services. But Fu’s warning was that the gap is widening: AI adoption runs at roughly 15% in developing countries against 26% in developed ones, and below 5% in low-income countries once China and India are set aside. The exclusion is layered, she argued, first in access to the technology, then in the digital literacy to use it, so the same communities fall behind twice.

Who owns the data

Vijay Modi of Columbia pushed the argument from access to ownership. He warned that AI can become “just another layer of dependency on large corporations,” and gave a concrete case: in Uganda, foreign vendors sold crop predictions from satellite data with no ground-truth, getting smallholders wrong and leaving them out, until the country sent young people out to interview 80,000 farmers directly. The lesson he drew was that the terms of AI are set by whoever collects and owns the data. Speakers for communities discriminated on the basis of work and descent made the sharper version: because data is the raw material of AI, groups that are missing from the data are written out of the systems built on it.

Vijay Modi, Columbia University: “Who owns the training data is actually going to be extremely important, and I think countries should watch out for this.”

The counting angle

This is the thread SDGCounting exists to follow, and the side event made it unusually direct. If a community is undercounted in the statistics, it is undercounted in the training data, and an AI system will treat it as if it were not there. The remedy the panel proposed is a measurement one: invest in disaggregated data by caste, descent, indigeneity, migration status, disability and gender, and gather it with the affected communities rather than about them. The Philippines offered the cleanest framing of what success should be measured against.

Deputy Permanent Representative of the Philippines: “True innovation is not measured by the sophistication of algorithms, but by the breadth of their benefits. A metric for successful AI can be how effectively it uplifts the marginalized, reaches the underserved, and empowers the vulnerable.”

Why it matters for the SDGs

The event maps most directly to SDG 10 (reduced inequalities), with the digital and AI divide and algorithmic bias in hiring, credit and social protection, and to SDG 9 (connectivity and compute), SDG 16 (governance and accountability) and SDG 4 (AI literacy). It sat deliberately downstream of the Global Dialogue on AI Governance in Geneva, whose closing line was replayed in the room, and of the Independent International Scientific Panel on AI. Where those set the global rules, this session asked the quieter question underneath them: whether the people the rules are meant to protect are even in the data.

Global Dialogue on AI Governance (video message): “The choice before us is not between facing AI or fear of it. It is between governing by design and drifting by default.”

Watch & read

UN Web TV, recording of the AI for Social Inclusion side event (8 July 2026).
UN Global Dialogue on AI Governance, the Geneva process this event drew on.
Independent International Scientific Panel on AI, the UN’s AI evidence base.
The Sustainable Development Goals Report 2026 · Full HLPF 2026 coverage.

Quotations are lightly edited from an automated (Otter.ai) transcript of the UN Web TV recording and should be read as close paraphrase; names and titles were reconciled to public records and reflect roles at the time. The many audience contributors, and one panellist’s personal name, were garbled in the transcript, so some speakers are cited by role or organization.