Three days ago, Salesforce and Anthropic gave their partnership a name: Claudeforce. The headlines focused on the tie-up between the number one AI CRM and Claude. For a security leader, though, the real story sits underneath. Your choice of AI model has quietly become a security decision.
What Claudeforce actually is
Claudeforce makes Claude a core reasoning model across Agentforce, and brings Salesforce into Claude as well. Claude runs through Amazon Bedrock, inside what Salesforce calls its Trust Boundary. In plain terms, the model reasons over your data without the data leaving Salesforce’s protected environment. Anthropic is the first model provider to sit fully inside that boundary.
Why model choice is now a security decision
For years, picking a model was a performance question. Which one is smartest? Claudeforce reframes it. When an agent can act on your CRM, the model is not just answering questions. It is touching regulated data and taking real actions. So the questions that matter become security questions. Where does the data live? Who can see it? What gets logged? Salesforce aims this setup at regulated industries: financial services, healthcare, cybersecurity. Those are the places where the answers matter most.
Where the model actually runs
The permissions trap
Here is the part that deserves a second read. Salesforce says Claudeforce needs no new permissions model. Your existing sharing settings carry over, and every user gets access on day one. That sounds convenient, and it is. But it also means your current sharing model becomes your AI governance model overnight. If permissions have drifted over the years, an agent will happily inherit that drift. Worth reviewing before you switch it on, not after.
Questions to ask first
So before you turn on agentic AI, a few questions are worth answering honestly. Where does sensitive data flow, and does it stay inside the boundary? Who reviewed the permissions the agents will inherit? What gets logged, and who watches it? Where does a human stay in the loop? None of these are exotic. They are the questions a good security leader always asks. Now they point at a powerful new actor.
A governance opportunity, not just a risk
It would be easy to read this as a warning. It is not. A trusted model inside a trusted boundary is good news for regulated teams. It just rewards the ones who treat rollout as a governance project, not a switch to flip. Get the permissions, logging, and data boundaries right, and agentic AI becomes an asset you can defend.
We’re solutions people
That is the work we do as a security partner. We are solutions people, so every engagement is built for how your business actually operates, never a template forced onto you. We review the permissions your agents will inherit. We set the guardrails, and put governance in before you scale, not after. As your partner, we own the outcome with you.