AI SECURITY / VISUAL RESEARCH
AI Security
Mind Map.
From model behavior to the boundaries that protect real systems.
My visual synthesis of Netskope AI security training notes, OWASP risk categories, and practical controls for tool-connected AI. Explore the layers, follow the attack paths, and inspect the defenses.
WHAT IT CONNECTS
Threats to controls.
The five-layer AI stack and the responsibilities that follow each deployment model.
Prompt injection, jailbreaks, supply-chain attacks, and MCP privilege boundaries.
Least privilege, sandboxing, content inspection, human oversight, and OWASP risk categories.
CONTEXT & SOURCES
Learning, made visible.
Created by Jordan L. Warren from personal Netskope certification study notes, supplemented by public technical documentation. This independent study map is not an official Netskope publication or endorsement. The five-layer stack is a teaching model; real threats can cross layers. Configuration examples must be checked against the deployed product version.
References: Claude Code security · Gemini CLI policies · MCP authorization · OWASP GenAI · OWASP Web Top 10:2025
Published November 2026. LLM category references on the board use the 2025 edition; agentic categories use the 2026 edition.