Digital Public Goods Alliance (DPGA)

A successful first Global Dialogue would achieve a clear international commitment to AI equity and move from principles to practice. It should outline relevant action areas and provide an actionable roadmap with concrete steps toward AI democratisation while addressing risks and aligning global rules for AI governance. Such a roadmap should mention digital public goods as a means to advance AI democratisation. Democratisation means going beyond access to AI technologies by enabling nations and communities to chart their own digital futures; it includes the ability to develop local solutions or adapt existing technologies to local contexts and ensures community agency in AI development and governance. To this end, the Global Dialogue should move beyond high-level principles to concrete mechanisms that ensure countries of the Global Majority have the infrastructure, resources and agency to co-create AI technologies rather than being mere consumers, and reconfirm the commitment of the Global Digital Compact to digital public goods as a defined concept to support solutions (AI systems, open data and open source software which are foundational for local, responsible AI development) that advance the SDGs.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

These priorities are inextricably linked to the DPG approach. Open source software, open data, and open AI models are vehicles for addressing power imbalances and dependencies, and for ensuring AI serves the public interest, enabling accessibility, local adaptation, and auditability for trust and safety while reducing costs and dependence on a few dominant tech entities. AI capacity-building is essential to ensure that this openness translates into actual utility; without the local technical skills in the public and private sectors to deploy, maintain, and govern these tools, the AI divide will only widen. Furthermore, focusing on the social, economic, and linguistic implications ensures that AI development respects cultural contexts and serves marginalised communities. This discussion stream should also address practices of digital extractivism, including data flows, resources (water, energy, raw materials), and land displacement in the course of expanding data centre infrastructure. Finally, transparency and accountability are the safeguards required to manage the risks of open systems, ensuring they remain safe, secure, and under human oversight - aspects which are reflected in the DPG Standard.

One critical issue is the contextual readiness of AI in Global Majority contexts. Current AI development often ignores the "in the wild" realities of low-resource environments, such as limited connectivity and access to compute, human and institutional capacity constraints and data scarcity. To address these challenges, the Global Dialogue should explicitly address the AI divide and promote a different approach to AI development and deployment by moving beyond the dominating assumption that meaningful progress depends on matching the massive, energy-intensive computing scale of big tech. An alternative narrative is frugal AI: smaller, specialised, and computationally efficient models trained on limited but highly relevant local data that can run on lower-end hardware and in resource-constrained environments. Such models are gaining significant competitiveness and are becoming viable for many real-world production tasks. Global Majority countries do not need the biggest models; they need the most efficient, governable, and cost-effective models that perform well in their domain, where fit-for-context, auditability and cost control are not optional. Additionally, the impact on the global labour force and the responsibility of big tech companies therein needs to be explicitly addressed, including exploitative (data) practices in the AI supply chain and the disruptive impacts of AI on labour markets. Moreover, the ecological impacts of AI (resource depletion and pollution) and economic disparities must be addressed. Again, a frugal AI approach is currently underappreciated and can reduce AI's carbon footprint. Lastly, safeguarding information integrity must be addressed, including the need for a negotiated declaration establishing expectations for content authenticity, electoral safeguards and non-interference, and human oversight in communication environments.

Challenges: The prevalence of "black box" algorithms creates a lack of trust and prevents local auditing and adaptation. AI hype often obscures the fact that many tools are not ready for diverse cultural, linguistic and "in the wild" contexts. The one-size-fits-all approach of LLMs is ill-fitted, economically and ecologically too costly, and lacks contextual relevance. There is a significant power imbalance in which a few corporations set the standards, leaving little room for local agency or cultural alignment. They also lead to lock-in effects and inflexibility due to mounting switching costs and dependencies. Opportunities: DPGs offer a proven model for designing open, trustworthy building blocks for digital transformation vetted against the DPG Standard. While no country can build its own sovereign AI stack, DPGs enable like-minded partners to collaborate to address dependencies and build transparent, accountable and locally developed technology on their own terms. This includes developing shared products and adapting existing solutions to local contexts to reduce costs. By promoting DPGs and open source AI, we can foster a global commons that lowers entry barriers for startups and governments in emerging economies, while building for inclusivity, innovation, transparency, trust and safety, turning AI from a source of inequality into a tool for sustainable development with local agency at its core.

The Global Dialogue should directly connect with the Digital Public Goods Alliance (DPGA). The DPGA stewards the DPG Standard, a community-governed benchmark for open and trusted software, data sets and AI systems. With the upcoming DPG4AI collection, we are building a hub that makes resources essential for responsible AI development widely available, helping to close the AI divide. These resources include open data and open source software for trust and safety testing, model development, and other aspects of the AI lifecycle, as well as the orchestration of agentic AI and AI systems. This collection will be expanded over time with members and partners such as UNICC, Mozilla, Current AI, the Open Data Institute, Digital Futures Lab, Open Future, ADB, national governments and local stakeholder groups. By working with the DPGA as a global collaboration platform and tapping into our work, the Global Dialogue can add value by ensuring that the commitments of the Global Digital Compact to advance DPGs for sustainable development gain recognition and feed into national AI strategies.

The DPG Standard serves as a prime example: It promotes governance through transparency and community-led development. By requiring AI projects to meet specific criteria for openness and "do no harm" principles, the DPG approach provides a concrete solution to the challenge of proprietary silos. It ensures that AI technologies remain inclusive, adaptable, and accountable to the communities and public they serve. Moreover, the DPG Standard promotes best practices and interoperability, such as open standards like model context protocols, which are essential for building an open AI ecosystem that caters to the needs of the Global Majority, including auditability, cost control and frictionless deployment. The United Nations initiated the DPG definition through the High-Level Roadmap for Digital Cooperation in 2019, which reflects a broad global consensus - essential for bridging disparate positions. We also propose learning from the open source software movement and encourage the Global Dialogue to lead the conversation on much-needed public-sector-led investment in research, safety, and trusted data infrastructures that fill market gaps and move us towards a global AI ecosystem that empowers everyone to build and own AI on their own terms.

Make a call for proposals to include community-led sessions during the Global Dialogue. From the DPGA's side, we would be pleased to co-chair a breakout session with a member state on the role of DPGs and open-source AI in bridging the AI divide as part of the Global Dialogue in July.

Global Majority voices broadly speaking, disadvantaged communities, children and the elderly. Open source developers and developers of digital public goods, and their supporters.

n/a

The DPGA promotes AI governance by championing DPGs-open source software, data, and AI systems-as a foundational framework for transparency and equity. By shifting away from "black-box" proprietary systems, the DPG approach offers a concrete path toward accountable AI that empowers national sovereignty and caters for local contexts. A primary tool in this effort is the DPG Standard, which includes specific criteria for AI systems as a category of digital public goods. This framework requires developers to provide all main elements of a system, including code, data and model weights, with a license that complies with the open definition or OSI's list of approved licenses, bias mitigation strategies, and privacy safeguards, ensuring that any AI that wants to be considered a digital public good is inherently trustworthy and does no harm by design. Furthermore, by facilitating a global registry of vetted solutions, the DPGA enables countries to discover and deploy tools that are already aligned with international ethical standards. This model addresses key challenges, such as the digital divide and algorithmic opacity, by democratising access to high-quality AI development resources. When AI models and datasets are open, they enable local adaptation and innovation, given capacity and skills, thereby addressing the monopolisation of technology. Through the DPG Standard as a community-driven benchmark and "standard of standards", the DPGA demonstrates that effective transparency and auditability are best achieved by embedding openness, interoperability, and fundamental rights directly into the technical architecture of AI systems, ultimately ensuring these technologies serve the collective public interest.