As the global conversation around making AI useful for everyone shifts from a “the bigger, the better” narrative to small, purpose-built, local and open AI, a critical question emerges: How do we ensure that AI tools truly serve people and the public good, rather than just a few concentrated interests?

At MozFest, we are excited to officially launch the DPGs for AI (DPG4AI) Collection—a curated portfolio of DPGs designed to drive responsible, equitable, and localised AI development.

Co-stewarded by the Digital Public Goods Alliance Secretariat, the United Nations International Computing Centre (UNICC), the Open Data Institute (ODI), and the Digital Futures Lab, this initiative translates global AI commitments—like those of the UN Global Digital Compact and the UN High-Level Advisory Body on AI—into a practical collection of software, data, content and AI model DPGs that advance the responsible development and deployment of AI systems. In a first step, these include trusted data infrastructure, semantic parsers, language datasets, safety tooling and a model comparison platform.

Thoughtfulness First: Getting the Foundation Right

When building a collection intended to serve all builders and adopters, no matter their geography, speed cannot come at the expense of intentionality. From inception, our primary goal was not simply to assemble a long list of open solutions, but to make sure we got it right.

In collaboration with our co-stewards and informed by interviews with ecosystem stakeholders, we spent significant effort developing rigorous collection criteria built on the DPG Standard.

We deliberately prioritised tools that:

• Target low-resource contexts & underrepresented domains: Ensuring solutions are adaptable for local developers, governments, and civil society operating under resource constraints.
• Have relevance at at least one stage of the AI lifecycle: Targeting open-source software, open training/fine-tuning/benchmarking datasets, open AI models, and open content collections relevant at different stages of the AI lifecycle.
• Demonstrate documented impact & safety: Grounding inclusion into the collection in real-world application and best practices for safety by design.

The Excitement Ahead: A Growing Living Repository

While getting the initial foundation right was essential, the launch at MozFest is just the beginning.

The open-source AI ecosystem is expanding rapidly, driven by the need for more accountability, cost control and local relevance of AI systems. We're also seeing major advances in edge computing, federated learning, localised dataset creation for training and benchmarking, and frugal AI frameworks. While we take a considerate, slower approach and do not want to push AI systems where they don’t belong and don’t deliver value for people, we plan to evolve the DPG4AI collection alongside these developments.

We are thrilled to watch this repository continue to grow on a rolling basis as new tools, models and datasets are submitted and vetted, and we are happy to work closely with DPGA members and ecosystem partners on this, such as CurrentAI, Mozilla, GIZ and UNDP.

A Call to Action

Whether you represent a project team working on open-source AI, a funder looking for high-impact tools to support, or a policymaker seeking trusted solutions:

• Explore the collection: Visit the collection page to browse the initial list of curated tools.

• Help us grow and submit your DPG: Have an open-source AI project, model, or dataset driving sustainable development? Check the DPG Standard and collection criteria and apply to be included.

• Finance DPG4AI: Use the collection as a curated pipeline to direct funding toward vetted, high-impact public AI components.

Together, we can build an AI ecosystem where innovation is open, benefits are shared, and technology is centred around people and their needs.