Salesforce Engineering outlines how they built a unified governance platform for enterprise AI agents, addressing the fragmentation that occurs when identity, data access, and API security are enforced in silos. The approach propagates user and agent identities consistently across multi-system workflows using Principle of Least Privilege token scoping. A centralized enforcement layer in Data 360 applies role-based, attribute-based, object, field, and row-level security plus dynamic masking to every data request regardless of origin. MuleSoft provides unified API policy enforcement and an AI Gateway for governing LLM access with intelligent routing and cost controls. Informatica adds automated data discovery, classification, and lineage mapping to eliminate governance blind spots. The Agentforce Trust Layer prevents external LLM providers from retaining customer data and masks sensitive information in responses. All trace data is aggregated for end-to-end auditability and anomaly detection.
Sort: