At the GPAI Summit 2023, a session on the Data Governance Working Group was organized on the third and final day of the event. Various teams presented their projects on data governance and AI to the audience.

The Data Governance Working Group collates evidence, shapes research, undertakes applied AI projects and provides expertise on data governance to promote data for AI being collected, used, shared, archived and deleted in ways that are consistent with human rights, inclusion, diversity, innovation, economic grown and societal benefit, while seeking to address the UN Sustainable Development Goals. 

The topics that the projects dealt with are mentioned below:

• From co-generated data to generative AI
• The role of the Government as a provider of data for AI
• Privacy-enhancing technologies
• Advancing research and practices for data justice
• Enabling data sharing through data institutions

From co-generated data to generative AI

Christiane Wendehorst and Kyoto Yoshinaga presented the subject. The objectives of their project were to examine concepts of co-generation, determine the extent to which existing concepts and regimes can be applied to the co-generation of content, examine how concepts developed for individual rights translated to collective digital rights and contribute to the transnational convergence of national and regional policies. 

In 2023, they conducted a literature review that produced a legal framework for analyzing and examined six co-generated scenarios using the framework. They found that big tech companies dominate the current landscape. A harm-led approach to understanding the impact is critical for a robust analysis.

The role of the Government as a provider of data for AI

Ching-Yi Liu and Jhalak Kakkar presented the report. The primary objective of their project was to guide governments in sharing government data with AI developers. They also provide legal and ethical principles that guide governments on responsible data sharing. They studied National Health, the Services, The Health Passbook, The Rapid Response Register for cash transfers in Nigeria and the Aclimate Agricultural Data Platform to attain these objectives. Based on their project, they recommended a set of principles for data sharing by governments.

Privacy-enhancing technologies

Shameek Kundu presented the report. From their project, they found that PET-enabled contact information helped create a temporal network that was nearly identical to the original. They learn that PET systems must be designed to balance data owners' security and privacy concerns. Efforts to specify and minimize the data field involved can improve the performance of the PET-enabled pandemic model. Education and early involvement of stakeholders is a necessity. 

Advancing research and practices for data justice

Allison Gillwald presented the report. The project aims to establish a framework for data justice that bridges the gap between research and practice, incorporating social and economic justice considerations in utilizing data for AI development. 

Enabling data sharing through data institutions

Thomas Nkoudou presented the report. The objective of their project was to map the data ecosystem and co-design a framework for trustworthy data exchanges. This year, they conducted fieldwork in Cameroon to scan the local data ecosystem to understand the key stakeholders, dynamics between data institutions and affected communities and gaps and challenges data governance faces. They also co-designed a framework allowing data institutions to develop safe, fair and equitable data sharing.