This past month, the Open Source Initiative (OSI) participated in the Digital Public Goods Alliance’s fourth Annual Members Meeting (AMM) in Brasília, an energizing gathering hosted in partnership with the Government of Brazil. The event marked the first AMM held in South America and brought together more than a hundred representatives from governments, multilateral organizations, NGOs, DPG product owners, and Open Source communities to align on the future of Digital Public Goods (DPGs) and Digital Public Infrastructure (DPIs).

Representing OSI were Katie Steen-James, Senior US Policy Manager, and Nick Vidal, Community Manager, who spent the week engaging with partners, facilitating discussions, and strengthening global momentum for open, trustworthy, public-interest technology.

Data Governance & Public Interest AI

A highlight of OSI’s participation was co-organizing and facilitating the roundtable “Data Governance & Public Interest AI”, alongside GeoPrism Registry, being represented by Nathan McEachen. The session built on a year-long, multi-stakeholder process, including OSI’s Deep Dive: Data Governance and DPGA’s Collaborative Action on Open Data for Public Interest AI.

Discussions explored six core challenges:

• Interoperability & technical barriers: Communities (particularly under-represented and under-resourced ones) face significant obstacles in accessing and deploying AI systems. Participants emphasized sandboxing environments to safely experiment, sector-specific interoperability strategies, and unified access points for users to simplify engagement. The discussion reinforced that reducing technical complexity is key to democratizing AI use and development at the local level.
• Cross-border collaboration amid geopolitics: The group examined how cultural norms, revenue models, and differing incentive structures shape international cooperation. They highlighted the importance of incentives related to climate change and other global challenges that transcend borders. Participants noted that political will varies widely, and collaboration must account for deep cultural and organizational differences. Balancing global ambition with geopolitical realities remains a central tension.
• AI’s environmental & climate impact: Conversations focused on energy consumption, the lack of transparency in AI-related emissions, and the difficulty of tracking environmental impacts without disclosure policies. Participants discussed pushing utilities for data center energy reporting. Hardware and software optimization, targeted AI use cases, and climate-resilience applications (such as landscape mapping and energy optimization) were highlighted as essential mitigation strategies.
• Bias, diversity & real-world harms: The group stressed that bias is multidimensional, emerging from data sources, problem prioritization, and entrenched resource allocation patterns. Technical tools to explore dataset biases, improved use of existing data, and broader inclusion in problem framing were identified as priorities. Participants also noted that development agencies and donors shape which solutions get built, often reinforcing inequities. Narratives and community needs shift over time, requiring adaptive governance approaches.
• Transparency, privacy & security tradeoffs: Participants explored competing definitions of transparency (from open data requirements to model openness) and how each interacts with privacy and security constraints. While open data can improve explainability and cross-border collaboration, it can also amplify bias or expose sensitive information. The group emphasized the role of Open Source security practices and collaborative approaches to identifying inherent biases. However, key questions remain about the performance impact of relying solely on open datasets.
• Openness, fair use, copyright & community compensation: This group grappled with the difficulty of attribution in AI-generated content and the imbalance in current commercial models, which reward aggregators over original data contributors. They raised concerns about the dominance of Global North datasets and how this shapes model behavior and cultural bias. While AI increases access to information, it also raises questions about copyright and equitable benefit-sharing. Proposed solutions included global legal frameworks, digital embassies to support data sovereignty, and the transparency advantages of open models.

DPGA and the future ahead for DPGs and Open Source

The DPGA’s Annual Members Meeting highlighted several priorities that resonate strongly with OSI’s mission, including promoting Open Source software, advancing public-interest AI, and strengthening global collaboration.

The city of Brasilia and its beautiful architecture were born from the audacious belief that “50 years of progress could be achieved in just 5.” That same audacity now echoes through the DPGA’s 50-in-5 campaign, a call for 50 countries to accelerate decades of digital progress into 5 years.

The conversations, commitments, and collaborations forged during this event carried the energy of a community ready to build something bigger than any one government or organization in isolation: an open, equitable, and resilient digital foundation for the world. Just as Brasília proved what is possible when courage meets collaboration, the DPGA community leaves this gathering ready to turn today’s aspirations into tomorrow’s shared Open Source future.