Does Salesforce AI Send Data to OpenAI?
This page documents the current public DealScope guide for Salesforce AI privacy and external-model questions.
The Current Public Answer
The public guide says the useful question is not only whether data is sent.
The more useful question is:
- what exact version of the data reaches the external model
- what Salesforce masking changes first
- what still needs to be verified in the org
What the Guide Emphasizes
The current public guide highlights Salesforce Trust Layer style masking as the key practical boundary before the model call.
It says that:
- names, emails, and phone numbers can be transformed before they leave the Salesforce side
- the payload seen by the model may differ materially from the raw record
What Teams Should Still Verify
The current public guidance says teams should still test their own workflow using known sample data.
The guide recommends verifying:
- what the model actually receives
- whether the workflow boundary matches the team’s assumptions
- how masking behaves in the actual configuration