Problem Statement:

Banks worldwide are currently looking to adopt AI in a compliant and scalable manner. However, many have not developed the appropriate controls. FINOS sees this as an opportunity to provide value to members by creating LLM controls for banks to start with.

Solution:

We'd like to have a joint scenario that enables a bank to see the following for an RAG LLM scenario:

• a reference architecture (provided by the AI group)
• a set of governance principles/framework (provided by the AI group)
• a target set of services
• a controls catalog for each of the services
• a definition of evidence and testing required
• a method for providing compliance evidence (provided by CFI)

Acronyms:
RAG - Retrieval Augmented Generation
LLM - Large Language Model