People should not have to stitch together five policy pages to get one answer.
Choose a product and generate a calm, useful ask: support email, public post, or procurement questions. The point is clearer answers, not louder theatre.
Email or support ticket
For people who want a calm, serious answer from the provider.
Hello Salesforce team,
I am trying to understand the public data-use posture for Salesforce Agentforce.
Could you publish or point me to one plain-language disclosure that clearly states:
- whether prompts, files, chats, or other user content are used to train or improve AI models
- whether the answer depends on user or administrator settings
- how to opt out where applicable
- what is said about retention and human review
The Production AI Institute's AI Data Use Index currently records the public answer as: "Salesforce says Agentforce runs through the Einstein Trust Layer with zero data retention agreements for third-party LLMs, and Salesforce does not currently use customer data to train generative AI models without an express opt-in."
A clearer single disclosure would help users, buyers, and reviewers understand the product without stitching together multiple policy pages.
Thank you.
Public post
For people who want to ask in public without sounding unserious.
I want a clearer public answer from Salesforce about how Salesforce Agentforce uses people's data for AI training, opt-outs, retention, and human review.
Current public summary from the PAI AI Data Use Index:
"Salesforce says Agentforce runs through the Einstein Trust Layer with zero data retention agreements for third-party LLMs, and Salesforce does not currently use customer data to train generative AI models without an express opt-in."
People should not have to guess.
https://www.productionai.institute/ai-data-use/salesforce-agentforce
Buyer questions
For organisations that need procurement-grade clarity before approval.
Ask Salesforce to answer these questions for Salesforce Agentforce:
1. What exact customer or user data categories may be used to train or improve AI models?
2. Which defaults apply by plan, region, or account type?
3. Which settings change the answer, and who controls them?
4. What retention applies to prompts, files, outputs, and metadata?
5. When can humans review the content?
6. Where is the single public disclosure a buyer can rely on?
PAI reference: https://www.productionai.institute/ai-data-use/salesforce-agentforce