Darryl Strucko

A Posture Orchestration Layer for AI/LLM Assets

Problem

Many tools address a portion of AI security, such as AI Security Brokers (AISB), model-hub scanners and agent permission controls. AI Security Posture Management (AISPM) looks at the whole AI estate: it collects their findings into one inventory, scores overall posture, and walks each finding through a recommended fix.

Approach

•Inventories every AI asset (models, agents, vector stores, gateways) in one AI-BOM and derives a single posture score from open findings.

•Five live use cases: shadow AI egress, RAG/memory poisoning, excessive agent agency, model drift, and model supply chain.

•Each finding carries its reporting source and inspectable evidence (simulated data), not a canned alert.

•Maps every finding to the OWASP LLM/Agentic Top 10 and MITRE ATLAS.

•Converts each finding into a recommended remediation playbook, then executes it with a full audit trail.

•Includes a real, read-only Hugging Face Hub lookup that checks a model's metadata and scan status without downloading it.