Salesforce Decode
Salesforcedecode
Back to questions
AgentforceIntermediatetrust-layerdata-maskingauditeinstein

Explain Einstein Trust Layer implications for Agentforce architects

Real World Scenario

Your CISO asks how customer PII is protected when prompts are sent to the LLM and whether data is used to train public models.

Expected Answer

• Einstein Trust Layer masks sensitive fields before LLM processing based on configured policies • Data is not used to train public foundation models—Salesforce contractually isolates tenant data • Audit trails capture prompt/response metadata for compliance review • Architects must still minimize PII in prompts—masking is not substitute for data minimization • Regional data residency requirements may affect which Einstein features are available • Supplement Trust Layer with org-level DLP policies and field classification programs • Document data flows for security questionnaires: CRM → Trust Layer → LLM → response path

Follow-Up Questions & Answers

Click to expand — each follow-up includes a direct, interview-ready answer

Direct answer: Einstein Trust Layer masks sensitive fields before LLM processing based on configured policies Also consider: Data is not used to train public foundation models—Salesforce contractually isolates tenant data In practice: Audit trails capture prompt/response metadata for compliance review Balance speed of delivery with maintainability.

Architect Perspective

Trust Layer is necessary but not sufficient—architects own data minimization and action-level authorization. Security reviews should see architecture diagrams, not marketing slides. Proactively document what leaves the org boundary.