The 5 Mistral AI security checks Black Cat runs
Black Cat SSPM evaluates 5 security policies against your Mistral AI configuration on every scan, classifies each finding by risk, and provides remediation steps. Below is the full list, grouped by category.
configuration
mediumAgent Uses Code Interpreter
Review agents granted the code-interpreter (arbitrary code) tool
highAgent References Org-Shared Library
Re-scope libraries an agent exposes that are shared org-wide
lowAgent Deployed Without Guardrails
Add guardrails to chat-deployed agents
sharing
highLibrary Shared Org-Wide With Documents
Restrict org-wide access to libraries that hold documents
mediumLibrary Shared Org-Wide Editable
Downgrade org-wide Editor access to Viewer or scope it