Security checklist for LLM deployments
A simple checklist enterprises can use to evaluate and harden LLM AI systems.
Security essentials
- RBAC and least privilege access
- Data boundaries for retrieval (permissioned RAG)
- Audit logs and traceability
- Prompt and tool-use policies
- Monitoring: latency, cost, quality, drift
Governance that scales
Define owners, change control, and incident response. The goal is predictable, repeatable operations.