Data Masking

Data masking settings in your Salesforce org and Model Playground determine the privacy of sensitive information and how that data is surfaced in a prompt response. Org-level settings are enabled by default.

When you create or update a model configuration, enable data masking in the Model Playground prompt settings. When enabled, data masking prevents data from leaving the Einstein Trust Layer to third-party large language models during testing and masks all sensitive data in a prompt response. When you disable the setting, sensitive data, such as names, birth dates, and email addresses, are no longer masked during testing.

At the org-level, the Einstein Trust Layer detects and masks selected sensitive data for models deployed in production.