Actors around the world are turning to data commons to improve access to high-quality data for AI development while advancing more equitable, accountable, and collaborative approaches to data governance.

We define data commons as collaboratively governed ecosystems that pool, steward, and provide responsible access to shared data resources. Yet initiatives operating under the label vary considerably in their governance structures, access models, funding arrangements, and relationships with contributors and affected communities. At the same time, many initiatives that resemble data commons in practice do not describe themselves using the term. Distinguishing data commons from adjacent models such as open data platforms, data cooperatives, or AI data pools is therefore not always straightforward. 

What We Did 

To better understand this landscape, the Open Data Policy Lab created the Data Commons for Generative AI Repository, which currently contains 49 initiatives spanning health, science, language, culture, mobility, environment, and other domains. The repository was developed through desk research and built on insights from our data commons work over the last two years, including research projects, studios, and the New Commons Challenge. In assembling it, we adopted a deliberately broad approach, capturing mature commons, federated infrastructures, cooperative models, consortium-based initiatives, and commons in development. 

As the repository expanded, we needed a consistent way to assess how closely different initiatives aligned with the principles of a data commons. Drawing on our research and the characteristics identified through our work on data commons, including the principles outlined in our Blueprint to Unlock Data Commons for AI, we developed an assessment checklist. The checklist translates these characteristics into a set of elements that can be used to compare initiatives and evaluate how closely they align with the data commons model.

We reviewed each initiative against the checklist using desk research. We did not use the framework as a scoring system, and inclusion in the repository did not depend on meeting a fixed number of criteria. Nor did we expect every initiative to fully implement all six elements. Instead, we evaluated how initiatives approached each element and whether, taken together, they demonstrated characteristics consistent with a data commons. 

After removing initiatives that no longer appeared active, we excluded initiatives when we found no evidence that contributors, affected communities, or member institutions could influence stewardship, access rules, accountability processes, or future development. Recognizing that the field remains at an early stage, we retained some commons-in-development that had not yet fully implemented governance or accountability structures when there was credible evidence that they were actively building toward them. 

Our objective was to reflect the diversity of institutional arrangements emerging across sectors while maintaining meaningful boundaries around the concept of a data commons.

What We Developed: Six Elements of a Data Commons 

These elements represent characteristics commonly found in data commons and provide a practical framework for assessment:

• Public Purpose: Whether the initiative aims to supply data needed to address public problems and support public-interest AI;

• Contributor Benefits: Whether contributors receive meaningful value in exchange for providing access to data, expertise, labor, or other resources;

• Funding and Support: Whether the initiative has identifiable sources of financial, institutional, technical, or operational support to sustain the commons over time;

• Access Controls: Whether the initiative establishes clear rules and mechanisms governing who can access data, under what conditions, and for what purposes;

• Participatory Governance: Whether contributors, affected communities, members, or participating institutions can meaningfully influence stewardship, rules, standards, or future development;

• Accountability Mechanisms: Whether the initiative includes safeguards, oversight processes, transparency measures, or other mechanisms that support responsible data use.

What We Found 

Across the repository, the vast majority of initiatives were created to serve a public purpose, offered clear benefits to contributors, had identifiable funding or support, and included some form of access control. Participatory governance and accountability mechanisms showed greater variation and ultimately became the most important criteria for distinguishing data commons from adjacent models. 

The review process clarified several categories that should not automatically be treated as data commons. These included:

• Centralized institutional or government-run platforms with no evidence of shared governance, community authority, or contributor stewardship;

• Open datasets or one-time data publications lacking ongoing governance, stewardship, or accountability;

• Peer-production, crowdsourcing, or co-production initiatives where participants contribute data or labor without meaningful governance authority; and 

• Alliances, standards bodies, or coordination networks that support data sharing but do not directly steward a shared data resource.

Many excluded initiatives exhibited openness, collaboration, or stakeholder engagement, but lacked mechanisms for shared stewardship and decision-making. The presence of a shared data resource alone was not sufficient; governance authority also needed to be distributed beyond a single central actor.

Looking Forward 

This framework is not intended to provide an exhaustive answer to how a data commons should be constituted. It provides a practical way to examine emerging models, compare institutional approaches, and draw clearer distinctions within an evolving field. As data commons continue to emerge across sectors and geographies, the framework can help practitioners, researchers, policymakers, and funders better understand how different models approach governance, participation, accountability, and sustainability. 

In our next post, we apply this framework across the repository and explore what it reveals about the current landscape of data commons. 

***

To learn more about our broader data commons work, including the New Commons Incubator, visit newcommons.ai.  

Back to the Blog