Beyond Open Source: Operationalizing Artificial Intelligence as a Digital Public Good

The current trajectory of artificial intelligence presents a stark paradox. While foundational AI models promise unprecedented breakthroughs in climate modeling, global public health, epidemiological surveillance, and education, the compute, data, and technical expertise required to build and maintain these systems remain concentrated in a handful of corporate entities and wealthy nations.

Dr. Serge Stinckwich

If AI is to accelerate the United Nations Sustainable Development Goals (SDGs) rather than widen global inequalities, AI systems must be governed, developed, and deployed as Digital Public Goods (DPGs).

Moving AI from an aspirational public good to an operational, accessible framework requires confronting a fundamental reality: traditional DPG standards—originally crafted for non-AI open-source software and open datasets—are no longer sufficient for the AI era.

Why AI Requires a New Class of Public Goods

For years, the global digital community has relied on the Digital Public Goods Standard to define open-source software, open data, open AI models, and open standards that adhere to privacy and do no harm. However, AI systems are fundamentally different from traditional software:

• Resource Intensity: Training and running state-of-the-art AI models demand massive computational infrastructure, energy, and high-quality curated data.

• Dynamic Lifecycle: AI models are not static codebases; they drift, require continuous fine-tuning, and depend on shifting data distributions.

• Black-Box Governance & Risks: Open weights alone do not equal transparency. Without visibility into training datasets, synthetic data methodologies, alignment techniques, and safety guardrails, "open" AI models can inadvertently propagate bias or security risks.

To build true AI as Digital Public Goods (AIDPGs), the global community must look beyond simply releasing open-source model weights. A comprehensive public ecosystem that encompasses the full AI stack is necessary.

Core Pillars of an Actionable AI DPG Framework

Based on multi-stakeholder assessments across the UN system, governments, civil society, and academia—detailed in the recent UNU flagship report, AI Systems as Digital Public Goods—three priority areas must be addressed to turn AIDPGs into an implementable pathway:

1. Open and Representative Data Infrastructure

AI is only as equitable as the data behind it. Most foundation models are trained on internet-scale data dominated by High-Income Countries, leaving global South contexts underrepresented.

• Public Data Commons: Building sovereign, localized datasets that represent regional languages, indigenous knowledge, and local environmental contexts.

• Synthetic Data Practices: Leveraging privacy-preserving synthetic data to train robust AI models in sensitive domains like digital health and public administration where real-world data is scarce or privacy-restricted.

2. Democratized Compute and Shared Infrastructure

An open AI model is unusable to a rural health center or local university if they lack the hardware to run or fine-tune it.

• Public Compute Pools: Establishing regional and global compute federations dedicated to public-interest research and SDG applications.

• Model Efficiency & Downscaling: Prioritizing small, domain-specific models, and the involvment of citizen and stakeholders through participatory modelingthat can execute on resource-constrained local infrastructures.

3. Sustainable Governance and Maintenance

The open-source ecosystem suffers when maintainers are underfunded. AI amplifies this challenge due to ongoing compute costs and safety monitoring.

• Shared Stewardship: Institutional mechanisms to fund the continuous maintenance, red-teaming, and safety audits of public interest AI.

• Science-Policy Integration: Direct pathways linking computational modelsand complex-systems approaches to evidence-based decision-making for policymakers.

What This Means for Macau’s Startup Ecosystem

As Macau actively pushes forward with its "1+4" economic diversification strategy—focusing on Big Health, modern finance, high-tech, and cultural tourism—the local startup ecosystem stands at a unique strategic crossroads. For founders and innovators in Macau and the surrounding Guangdong-Macao In-Depth Cooperation Zone in Hengqin, thinking about AI through the DPG lens offers a distinct competitive edge:

• Focusing on Application and Fine-Tuning over Model Training: Macau startups do not need to burn millions of USD trying to build frontier foundation models from scratch. By leveraging open weights AI Model (e.g. from Mainland China) and open AI DPGs, local startups can focus capital on developing high-value, domain-specific vertical applications, fine-tuning open weights for local public health, smart tourism, and regional logistics.

• Capitalizing on the Lusophone and GBA Bridge: Macau sits as an ideal international gateway between the tech powerhouses of the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) and Portuguese-speaking countries. Local entrepreneurs can pioneer multilingual, cross-cultural AI datasets and translation tools as digital public goods, positioning Macau as a trusted testing ground for ethically governed, cross-border AI solutions.

• Building Privacy-First, Compliant Tech with Synthetic Data: Cross-border data transfer between Macau, Mainland China, and international markets presents complex regulatory dynamics. By adopting privacy-preserving synthetic data techniques championed by AI DPG frameworks, local startups can build and train robust algorithms for fintech and digital healthcare without violating strict data sovereignty laws.

• Unlocking Public Sector and International Procurement: Aligning early-stage technology with international the DPG standard makes Macau startups far more attractive for public-sector contracts, international development bank grants, and regional ESG-focused venture capital investments.

Championing Public-Interest AI: Dr. Serge Stinckwich

This roadmap for AI as Digital Public Goods is rooted in the long-standing research and advocacy of Dr. Serge Stinckwich, a computer scientist and Senior Research Fellow at United Nations University Office in Paris. With over sixteen years of experience working at the intersection of digital technology, complex systems modeling, and sustainable development across Asia, Africa, and Europe, Dr. Stinckwich has been a pioneer in harnessing open technology for global public policy.A central chapter of his global footprint was written in Macau.

From March 2020 through March 2026, Dr. Stinckwich served as Head of Research at the United Nations University Institute in Macau (UNU Macau). During his six-year tenure in Macau SAR, he built and led an interdisciplinary team dedicated to responsible artificial intelligence, gender and technology, digital health, social simulation and participatory modeling. His footprint in Macau established the territory as a key regional hub for science-policy dialogue, convening international researchers, local startup hubs, and diplomats to address the ethical and structural challenges of emerging technologies across the Indo-Pacific and global South.

As a founding member of Open Source United—the UN network advocating for open source principles—and a primary author behind flagship UN assessments (including the comprehensive AI Systems as Digital Public Goods report), Dr. Stinckwich continues to lead international initiatives bridging technical innovation with global equity.

This article was conceptualized and guided by Dr. Serge Stinckwich.

========================

Interested in joining the Macao Startup Club?

Become a member to connect with founders, innovators, and global startup ecosystems.

👉 Register here: https://www.link-card.xyz/associations/msc/membership