Salesforce wraps every agent interaction in what it calls the Einstein Trust Layer. Its themes are: permissions that carry through, so the agent sees only what the signed-in person can see, down to individual fields; zero data retention agreements with the model providers, so your data is not kept or used for training; detection of harmful responses; an audit trail of prompts, responses and actions; and data masking of sensitive values for certain features. Which of these applies depends on the feature in use, and Salesforce documents, for example, that masking does not apply to agents. We classify your data, scope the access, write the policy and test it before anything goes live, and we confirm which protections apply to your set-up.