OpenMined at the Datasphere Initiative Roundtable on Advancing Global AI Governance

The Datasphere Initiative has published a report summarising a roundtable event they hosted on the topic of “Advancing global AI governance: Exploring adaptive frameworks and the role of sandboxes”. The event, held on the fringes of the Paris AI Action Summit, brought together global leaders to discuss critical aspects of AI governance. OpenMined’s Dave Buckley was honoured to join distinguished speakers from Microsoft, the OECD, Singapore’s Infocomm Media Development Authority, the Global Center on AI Governance, and Japan’s Ministry of Internal Affairs and Communications, for a dynamic discussion focused on opportunities for global collaboration on AI governance.

During the discussions, Dave provided OpenMined’s perspective on the need to move beyond high-level frameworks and voluntary commitments for AI governance, and consider how we build the right technical tools that can help operationalise such frameworks, ensuring meaningful governance and oversight of AI systems in practice.

He argued that independent oversight of AI systems today is inhibited by information access challenges, for example: an inability to access training data can mean evaluators are unable to verify whether benchmark data is absent from the training data; evaluators may not have sufficient access to verify whether systems running in production are the same as those that underwent independent evaluation; a lack of access to usage data prevents researchers from understanding real-world harms of AI.

OpenMined has been building and deploying technical governance infrastructure that can help overcome these challenges. Last year, we collaborated with EleutherAI to demonstrate a privacy-preserving third-party safety audit of proprietary training data and with UK AISI and Anthropic to demonstrate third-party model evals with mutual secrecy using secure GPU enclaves.

We are excited to build on this work and scale our infrastructure to support governance use cases across the AI lifecycle. We build open-source technology with the belief that AI governance is a global challenge and so must be underpinned by infrastructure that serves as a global public good. We look forward to continued collaboration with global partners to advance this mission.

Interested in updates about this topic?

Sign up to be notified if similar content is posted

This field is for validation purposes and should be left unchanged.
Author:
Category:
policy
Continued Reading...
View all posts
Balancing Innovation and Privacy: Congress Eyes PETs as a Solution to AI Data Challenges in the AI Task Force Final Report
Response to Ofcom’s Online Safety Information Guidance Consultation

OpenMined is a 501(c)(3) non-profit foundation and a global community on a mission to create the public network for non-public information.

With your support, we can unlock the world’s insights while making privacy accessible to everyone.

We can do it, with your help.

Secure Donation

$
$
Philanthropist looking for more?
Contact us