Well-governed data availability and access for artificial intelligence: Demonstrating the practical use of privacy-enhancing (and adjacent) technologies

Objectives

Using AI algorithms to address multi-stakeholder public good problems (e.g. pandemic response, climate action) requires sourcing and sharing data in a secure way while respecting data privacy, sovereignty, access control, intellectual property (IP) rights, and preventing data misuse.

Privacy-enhancing technologies (PETs), and more generally, the emergent concept of “structured transparency” can help. Although there have been recent efforts to identify how PETs can potentially support real-life use cases, the deployment and adoption of such technologies remain relatively limited. This project aims to demonstrate the viability of these technologies in helping achieve the UN SDGs while addressing these data challenges.

Utilising Privacy-Enhancing Technologies to overcome data barriers for social good

In one of the world’s first cross-border collaborations on PETs, practical demonstrations for AI systems will be conducted, drawing insights to develop practical guidance for AI developers and owners of AI systems. The experience will guide future research and development, business adoption of PETs, and contribute towards developing international standards.

To test how these technologies can further enable the development of AI systems, the project will deliver

(1) a practical demonstration of how PETs can help improve data availability for AI use cases that are beneficial to humanity,
(2) practical guidance and framework(s) for data scientists and AI developers on how to work with such technologies,
(3) guidelines on further development and adoption of AI technologies (which could also translate to international standards),
and (4) an outreach plan that yields greater awareness of, and confidence in, technology solutions to address privacy, IP, and sovereignty concerns.

This project explores how to enable greater data availability to support innovation and improve competition for data/analytics-enabled products and services, all with public benefit. In general, more frictionless data sharing between organisations and/or countries will enable learning and innovation. It will also support smaller organisations or corporations to compete more effectively with large (and sometimes monopolistic) data-rich organisations that have access to massive datasets within their organisational boundaries.

Managing data transparently, effectively and responsibly has never been more important, especially when it comes to AI and machine learning. We need to ensure data is available and accessible so we can unlock its potential for humanity, and I'm delighted to see that this GPAI-led initiative is taking practical steps to demonstrate how Privacy-Enhancing Technology (PET) can help improve data availability. This guidance and framework are vital for AI developers and data scientists as we shape the future of data governance.

Joanna Shields

CEO of BenevolentAI, Previous Co-Chair of the GPAI Steering Committee and Chair of the GPAI Multistakeholder Experts Group Plenary

The GPAI-led project will show how these ideas work in an end-to-end system and will inform our future policies and practices. This demonstration project is just what the public and private sectors need to help shape the future of data governance.

Dewey Murdick

Director, Center for Security and Emerging Technology, Georgetown University and former GPAI Data Governance Working Group expert

Resources

Team

Stephanie King

CEIMIA

Director of AI Initiatives

Kim McGrail

University of British Columbia - UBC

GPAI expert and Project co-lead

Shameek Kundu

Infocomm Media Development Authority - IMDA, Singapore

GPAI expert and Project co-lead

Project Advisory Group

Marc Rotenberg

Bertrand Monthubert

Ching-Yi Liu

Andrea A. Jacobs

Christian Reimsbach

Michael O'Sullivan

Collaborative opportunities with CEIMIA

Collaborating with CEIMIA means contributing to the development of responsible AI. For this project we welcome three types of collaborators:

Organisations that have a possible AI-for-social-good use case who would be interested in exploring what a demonstration of PETs could look like for their application(s); and
Relating to our first demonstration use case – individuals and/or teams working at the intersection of incorporating location-based data into contact networks (or, even more specifically, pandemic modelling contact networks).

Additionally, we welcome expert representatives from external organisations in this domain to join the team as members of the Project Advisory Group.

Contact us