When an AI agent acts on your behalf, who is accountable if something goes wrong?

AI agents are beginning to make transactions, verify identities and take action without a person approving every step. This raises new questions for the systems we rely on to establish trust: How do we verify who, or what, is behind an agent? Who is responsible for its decisions? And how can trust standards work across borders when agents operate globally but regulation remains local?

This workshop brings together industry practitioners, an academic researcher and a global policy convenor to explore these questions through developments in commerce and compliance. Drawing on emerging approaches such as Know Your Agent frameworks, speakers will discuss where risks are appearing, how identity and accountability can be established, and what governance needs to look like as adoption grows.

The discussion will offer participants a clearer understanding of the emerging risks and the safeguards needed to build trust in the agentic economy.

Trusted cross-border data collaboration and analytics has useful applications in payments, addressing financial crime and helping central banks and financial authorities monitor systemic risk and enhance supervision. Privacy enhancing technologies have demonstrated good potential in enabling cross-institution data collaboration while preserving privacy, confidentiality and institutional control, but have failed to scale beyond pilots. This roundtable will examine emerging privacy-preserving technology applications in the financial sector and discuss the technological, governance and policy foundations required to scale trusted data collaboration.

How can AI agents interpret regulation reliably when people already struggle to keep pace with its volume and complexity? Trusted agentic AI requires authoritative regulatory information organised so that people and machines can understand and use it consistently.

This is fundamentally a human-and-machine model. Machines can analyse regulation systematically and at scale, connecting findings to evidence. People contribute legal and institutional context, exercise judgment and remain accountable for consequential decisions. Neither should function in isolation: uncertainty and disagreement must be visible and escalated for human review.

This workshop will explore the regulatory data foundations needed to support this partnership. Participants will consider how shared ways of classifying rules and defining regulatory concepts could enable agents to retrieve the correct provisions, validate interpretations and cite authoritative sources—while reducing duplication for public authorities and global banks.

The discussion will examine different approaches to organising and governing regulatory information, including whether shared standards can enable interoperable systems without requiring a single global classification framework or central database. It will also consider how to preserve jurisdictional authority and institutional control.

Participants will gain a clearer understanding of how to combine machine scale and consistency with human context, judgment and accountability.