Organizational Identity Research & Publications

Agentic AI in Payments: Establishing Interoperable Trust and Control


  • Date: 2026-08-13
  • Views:

Download as PDF

AI agents are starting to initiate and execute payments on behalf of businesses and consumers, and payment networks and technology providers have launched infrastructure to support them. For each payment, a counterparty needs to know who authorized the agent and what it was allowed to do.

This GLEIF working paper argues that existing law can already allocate responsibility for an agent’s actions. The harder problem is practical: proving what authority an agent was given, and keeping evidence that holds up if a payment is disputed.

Because a universal identity standard for continually retrained agents may take time to settle, the paper prioritizes verifiable authority. It explores how the Legal Entity Identifier (LEI) and verifiable LEI (vLEI) could support Agent Mandate Credentials. Through a chain of vLEI credentials, each agent would be linked to the authorized person who issued its mandate and, from there, to the organization’s LEI. A payment service provider could then computationally verify the mandate and its limits before executing a payment, without the agent holding the underlying account credentials.

The paper works through four use cases and outlines two pilot concepts. The use cases are supplier payments and treasury automation, payment for data in due diligence, cross-border B2B settlement, and multi-agent workflows.

Published in GLEIF’s Working Paper Series, the paper is intended to inform discussion and does not represent an official GLEIF position.