In the GenAI Connect rule, there is an option to mask sensitive data before sending it to the LLM. However, the same capability is not available in GenAI Coach and AI Agent rules. What is the recommended approach for handling PII and sensitive data masking in these cases? Should we implement custom masking logic within the prompts, data sources, and tools to ensure sensitive information is protected before being processed by GenAI Coach or AI Agents?
Masking just the prompts for agentic solution will be not enough because there is part of this agentic thing which is the harness/orchestration layer. So there you dont have a control on the masking so I would say masking at the source is the only option which might be troublesome, not sure if you have better option right now.
PII data masking for Agent will be available in 25.1.4. GenAI Coach - not sure at the moment.
I am in Pega 26.1.1. Could you guide me the configuration in Agent rule for PII data masking? Or something else required for this?
@Vaisakh.V.A This will be available in 26.1.2 for Agent rules to mask the PII data.