Security checks / Check reference / AI & Copilot
Apply Least Privilege For Agent Functions
High Severity Microsoft 365 security check in the AI & Copilot domain.
Apply least-privilege principle to AI agent functions to limit blast radius of misuse AI agent tools with excessive permissions can read, modify, or delete data beyond their intended scope if a prompt injection occurs Siemserva runs this check automatically across your tenant and maps every finding to the compliance frameworks below.
Check summary
What this check verifies
Apply least-privilege principle to AI agent functions to limit blast radius of misuse
Why it matters
AI agent tools with excessive permissions can read, modify, or delete data beyond their intended scope if a prompt injection occurs
How to fix it
Scope agent tool functions to minimum required permissions
Siemserva flags every tenant where this check fails and, with Senserva Trustworthy AI remediation, can propose a validated fix you approve before it applies.
Ask your own AI about this check
Copy this prompt into Claude, ChatGPT, or Copilot. The facts are included, sourced from this page.
Compliance framework mappings
Findings from this check serve as evidence for:
Related AI & Copilot checks
Common questions
What does the Apply Least Privilege For Agent Functions check verify?
Apply least-privilege principle to AI agent functions to limit blast radius of misuse It is a High Severity check in the AI & Copilot domain, one of the 650+ checks Siemserva runs across Microsoft 365, Intune, Defender, and Entra ID.
How do I fix a Apply Least Privilege For Agent Functions finding?
Scope agent tool functions to minimum required permissions Siemserva ranks the finding by Severity and, with Senserva Trustworthy AI, can propose a validated, approve-before-apply configuration fix.
Which compliance frameworks does this check evidence?
Findings from this check map to SCuBA, MCSB, NIST, SOC2, CIS, and MCSB controls GS-1, PA-1. That mapping is what turns a scan finding into audit evidence.
Can I ask my own AI about this check?
Yes. This page includes a free copy-paste AI prompt carrying the check's facts: what it verifies, the risk, the remediation, and the framework mappings. Paste it into Claude, ChatGPT, or Copilot; cite-worthy detail without giving an AI access to your tenant.
All 672 checks with severities and mappings: the Microsoft 365 security checks catalog.