Which part of the Einstein Trust Layer architecture leverages an organization's own data within a large language model (LLM) prompt to confidently return relevant and accurate responses?
Dynamic Grounding in the Einstein Trust Layer architecture ensures that large language model (LLM) prompts are enriched with organization-specific data (e.g., Salesforce records, Knowledge articles) to generate accurate and relevant responses. By dynamically injecting contextual data into prompts, it reduces hallucinations and aligns outputs with trusted business data.
Prompt Defense (A) focuses on blocking malicious inputs or prompt injections but does not enhance responses with organizational data.
Data Masking (B) redacts sensitive information but does not contribute to grounding responses in business context.
[Reference:, Salesforce Help Article: Einstein Trust Layer – Dynamic Grounding ("How Dynamic Grounding Works" section)., Einstein Trust Layer Technical Overview: "Contextual Accuracy with Dynamic Grounding.", ]
