Einstein Trust Layer

Question 1

Guardrails and closed loop system built into the Einstein 1platform to protect data

Answer 1

Einstein Trust Layer

Question 2

What does Einstein Trust Layer use to prevent sensitive data from reaching LLM?

Answer 2

Data Masking

Question 3

Where is trust layer turned on?

Answer 3

Data Masking in Einstein Trust Layer setup

Question 4

What is PII

Answer 4

Personally Identifiable Information

Question 5

After the LLM has processed your prompt, what prevents any data referenced from being stored by the LLM?

Answer 5

Zero Retention Policy

Question 6

What are the layers of the Einstein Trust Layer?

Answer 6

1. Prompt
2. Secure Data Retrieval
3. Dynamic Grounding, RAG
4. Data masking
5. Passed to LLM - LLM processes and sends back with zero data retention
6. Response
7. Toxicity Detection
8. Data Demasking
9. Audit trail and feedback

Question 7

How is audit data turned on?

Answer 7

Setup - Einstein Feedback

Question 8

What is the Trust Layer?

Answer 8

A sequence of gateways and retrieval mechanisms that together enable trusted and open generative AI.

Question 9

Why did Salesforce create the Einstein Trust Layer?

Answer 9

To provide a toolbox of features to protect customer and company data

Question 10

What is the #1 Salesforce value?

Answer 10

Trust

Question 11

Where is Prompt Defense available?

Answer 11

Prompt Builder, Prompt Template Connect API, Prompt Template Invocable Action

Question 12

What powers generative AI applications?

Answer 12

Prompts

Question 13

What creates a great prompt?

Answer 13

Clear instructions, contextual information, constraints

Question 14

What is a prompt template?

Answer 14

Instructions and placeholders that will be filled with business data.

Question 15

What is the process of replacing placeholder fields with page context, merge fields and relevant knowledge from a record?

Answer 15

Dynamic grounding

Question 16

What starts the process of dynamic grounding?

Answer 16

Secure data retrieval

Question 17

Can users access all data with secure data retrieval?

Answer 17

No. They can only access what their Salesforce permissions allow them to access.

Question 18

What uses machine learning and search methods to find relevant information in other data sources that can be included in a prompt?

Answer 18

Semantic retrieval

Question 19

What is the process of tokenizing prompt values to replace private information based on what a placeholder represents?

Answer 19

Data Masking

Question 20

What is prompt injection?

Answer 20

An attack where an attempt is made to get around restrictions and attempt to perform tasks or manipulate a model in a way it was not designed

Question 21

What is prompt defense?

Answer 21

An attempt to protect against attacks by inserting instructions to the LLM about what should or should not be addressed or generated.

Question 22

What policy means that no customer data including prompt text and generated responses can be stored outside of Salesforce?

Answer 22

Zero Data Retention

Question 23

What protects responses from hateful, violent, sexual, identifiable, physical or profane responses?

Answer 23

Toxic language detection

Question 24

What allows users to provide feedback to a response?

Answer 24

Feedback Framework