AI Risk Practitioner Track

Risk, governance, security & compliance professionals9 steps · ~68 min

Work through the OWASP-aligned threat landscape with hands-on incidents, mapping each to root causes, controls, and frameworks (OWASP, ATLAS, NIST AI RMF).

By the end: You can map a reported AI incident to the OWASP LLM Top 10 / MITRE ATLAS, name its root cause, and pick the controls that would have contained it.

0/5 scenarios completed

Start the path →

1
🗺️Tool-Using Agent
Reference architecture
Anchor on the tool-using agent — the workhorse architecture.
2
📧The Email That Gave Orders~10 min
Interactive scenario · intermediate
The 'lethal trifecta' in action: injection + tools + data.
3
🖼️Zero-Click Leak by Picture~10 min
Interactive scenario · advanced
Zero-click data exfiltration via auto-rendered output.
4
🔑The Agent With the Master Key~9 min
Interactive scenario · intermediate
Over-privileged agent, one destructive command.
5
🧠The Memory That Wouldn't Die~10 min
Interactive scenario · advanced
A one-time trick that persists across every session.
6
☠️Poisoning the Well~9 min
Interactive scenario · intermediate
Corrupt the knowledge base, corrupt every answer.
7
📰ForcedLeak — Salesforce Agentforce CRM exfiltration (CVSS 9.4, no CVE)
Real-world case · 25 Sep 2025
Read a disclosed real-world version: ForcedLeak in Salesforce Agentforce.
8
🧭Framework hubs
OWASP LLM Top 10 · MITRE ATLAS · NIST AI RMF
Walk the framework hubs — every OWASP / ATLAS ID with its real incidents.
9
🛡️Control Library
Preventive · detective · corrective
Finish in the Control Library: preventive / detective / corrective coverage.

Next path

Agentic & Multi-Agent Security →

Go deep on the hardest surface in modern AI: autonomous agents, tools, MCP integrations, and teams of agents coordinating.