Course

AI Security & Identity

Treat every agent as a principal with constrained authority.

AdvancedSecurity, IAM, and platform teams responsible for action-taking AI.Coming soon — self-paced, cohort, or hybrid

Outcomes

Model humans, agents, and workloads as distinct principals.
Design delegation and token exchange without confused-deputy paths.
Place authorization at tool execution, not only at login.
Threat-model prompt injection, RAG poisoning, and capability chaining.

Who It Is For

Security architects
IAM engineers
Platform and runtime owners

Prerequisites

Working knowledge of OAuth/OIDC
Identity & Authorization topic on Learn

Curriculum

Identity

Human, agent, workload identity
OIDC and OAuth
SPIFFE

Delegation

Token exchange
On-behalf-of
Short-lived credentials

Threats

Prompt injection
Confused deputy
Tool poisoning

Runtime policy

Capability boundaries
Audit
Continuous authorization

Architecture Labs

Labs focus on architecture decisions, not toy coding exercises.

Red-team an agent tool path
Design a PEP for tool dispatch

Included Resources

Threat model templatesChecklistsLabs

Instructor

Hammad Abbasi — AI & Security Architect. 16+ years across identity, distributed systems, and production agent architecture. Speaker at EIC KuppingerCole.