Einstein Trust Layer in Salesforce: Security for Safe Use of Generative AI

Salesforce has always prioritized the safety and reliability of customer data. In this context, the "Einstein Trust Layer" was developed to ensure data protection and trustworthiness in the use of generative AI. This blog explains in detail how the Einstein Trust Layer works and how AI can be leveraged while protecting customer data.

1. Overview of the Trust Layer

Einstein Trust Layer provides a set of security features to protect data and privacy when using generative AI. This layer ensures high-level security whenever AI accesses customer data across all Salesforce applications.

Overall process diagram of the Einstein Trust Layer

2. The Prompt Journey

The Trust Layer ensures the security of the process when a prompt is sent to generative AI. Specifically, the prompt goes through the following steps:

Dynamic Grounding: The prompt is enhanced with relevant organizational data to generate a contextually appropriate response.
Data Masking: Personal information within the prompt is masked before being sent to AI, ensuring that it is protected and does not leak externally.

Diagram showing how the prompt is processed

3. The Response Journey

Once the generative AI response is returned, it is processed safely again through the Trust Layer. This process includes the following key steps:

Zero Data Retention: To ensure no data is stored by external AI providers, the response data is immediately discarded.
Harmful Language Detection: The response is checked for harmful language before being sent to the customer.

Flowchart of how responses are processed through the Trust Layer

4. Guardrails for Safety and Reliability

The Trust Layer provides additional security guardrails to ensure that data is properly protected and that AI does not generate unintended responses. This includes protection against prompt injection attacks and a validation process to ensure that the generated responses are appropriate.

[Illustration Image Insertion Here]
Illustration: Diagram explaining the functions of guardrails

5. Conclusion and Future Outlook

Salesforce's Einstein Trust Layer is an innovative solution that prioritizes safety and reliability in the use of generative AI. With this layer, companies can confidently leverage AI to improve productivity. Going forward, Salesforce will continue to offer even more reliable AI solutions.

Diagram emphasizing reliability and safety

Through this blog post, we hope you gained an understanding of the importance and specific features of the Einstein Trust Layer. Salesforce will continue to provide reliable AI tools to support your business in the future.