Zero-data retention with third-party LLMs
Key Takeaway
Under zero-data retention, an external LLM provider doesn't retain the data Salesforce sends it and doesn't use that data for model training. The data is deleted after the response goes back to Salesforce. The policy still protects agent data, even though data masking is turned off for agents.
Zero-data retention is an Einstein Trust Layer policy that limits what external LLM providers, such as OpenAI, can do with the data Salesforce sends them. The policy is a key protection for Agentforce agents because data masking is turned off for agents. Without masking, sensitive values in an agent's prompt aren't replaced with placeholder text before the prompt is sent to the model.
Zero-data retention in the Einstein Trust Layer
Zero-data retention is one of the agreements in the Einstein Trust Layer. Salesforce partners with all its external LLM providers, such as OpenAI, Anthropic, and Google, to enforce it.
Einstein Trust Layer capabilities, including zero-data retention, apply only to generative AI and Agentforce features.