Back to AI Security

AI SecurityLivebeginner60 hours

AI Security Blue Team

Learn how to defend AI applications through structured lessons, practical reviews, and hands-on rooms that teach defenders where control really belongs in the stack.

Modules

5

Rooms

25

Catalog Status

Live

Path Outline

Modules and Rooms

Modules that are already published show their live rooms below. Modules still being restored stay visible as roadmap placeholders so learners can see the full path direction.

1

Defensive Foundations

Build the core blue-team mental model for AI systems by learning how layered defense, attack surface mapping, trust boundaries, and the prevention-detection-response loop fit together.

5 live rooms

Defense in Depth for AI

Learn the fundamentals of AI security defense and why strong blue teams secure the full application stack, not only the model.

beginner55 min90 XP

AI Attack Surface for Defenders

Learn how defenders map the real attack surface of AI systems: where untrusted input enters, where privilege exists, and where failures can become operational impact.

beginner60 min100 XP

Trust Boundaries in AI Systems

Learn where trust changes inside an AI system, why those crossings matter, and how defenders stop untrusted content from quietly becoming privileged action.

beginner60 min100 XP

Prevention, Detection, and Response

Learn how blue teams reduce AI incidents before they happen, notice them when prevention fails, and respond in a way that limits damage and improves the system.

beginner65 min110 XP

Topic Rewind Recap

Rewind the first Blue Team foundation block and consolidate the core ideas: layered defense, attack surface, trust boundaries, and the prevention-detection-response loop.

beginner35 min120 XP

2

Prompt and Context Defenses

Learn how defenders harden prompts, separate trusted policy from untrusted context, and reduce prompt-injection risk in real applications.

5 live rooms

Prompt Injection for Defenders

Learn what prompt injection is, how it reaches AI systems through direct and indirect channels, and why defenders treat it as a system-level security problem.

beginner60 min100 XP

Trusted Instructions vs Untrusted Content

Learn how AI applications mix system policy, user input, retrieved content, and tool results, and why defenders must keep trusted instructions separate from untrusted content.

beginner60 min105 XP

Prompt Leakage and Sensitive Information Disclosure

Learn how AI systems leak internal instructions and sensitive information, and how defenders reduce disclosure risk without breaking legitimate assistant workflows.

beginner65 min110 XP

Defensive Prompting, Output Controls, and Safe Fallbacks

Learn what prompt-layer defenses can realistically do, why output handling matters, and how blue teams design assistants that fail safely instead of turning model mistakes into incidents.

beginner65 min115 XP

Topic Rewind Recap

Rewind the prompt and context defense block and turn the core ideas into a review workflow: identify hostile language, separate trust levels, reduce disclosure risk, and fail safely.

beginner40 min125 XP

3

Retrieval and Tool Security

Secure RAG pipelines, classify document trust, scope tool permissions correctly, and add approval gates where the model can trigger real-world actions.

5 live rooms

RAG Security Fundamentals

Learn why retrieval changes the AI security model, how untrusted documents reach the model, and what defenders must review before treating retrieved content as useful context.

beginner60 min100 XP

Document Trust, Provenance, and Ingestion Hygiene

Learn how defenders decide which documents belong in a knowledge base, how provenance and metadata change trust, and why ingestion hygiene is one of the strongest RAG controls.

beginner60 min100 XP

Tool Calling and Excessive Agency

Learn why tool-enabled assistants are riskier than chat-only systems, what excessive agency means, and how defenders reduce harm through permissions, approvals, and narrow execution scope.

beginner65 min100 XP

Identity, Authorization, and Tenant Isolation

Learn how user identity, authorization, and tenant boundaries must survive the trip through retrieval and tool use so the assistant never acts beyond the user's real access.

beginner65 min100 XP

Topic Rewind Recap

Rewind the retrieval and tool security block and apply the core ideas in one defender workflow: classify document trust, preserve access scope, narrow tool permissions, and keep risky actions behind approval.

beginner45 min125 XP

4

Monitoring and Incident Response

Build the visibility, alerting, and response playbooks needed to detect abuse, triage AI incidents, and recover safely when controls fail.

5 live rooms

What to Log in AI Systems

Learn what telemetry defenders need from AI systems, why prompts alone are not enough, and how logging decisions shape detection and incident response later.

beginner60 min100 XP

Detection Engineering for AI Abuse

Learn how defenders turn AI telemetry into detections, what suspicious patterns look like in practice, and why detection logic must focus on behavior rather than one perfect keyword.

beginner60 min100 XP

AI Incident Triage and Containment

Learn how defenders triage suspicious AI incidents, decide what matters first, and contain unsafe behavior without creating more harm than the original issue.

beginner65 min100 XP

Recovery, Root Cause, and Lessons Learned

Learn how defenders move from containment to recovery, how root-cause analysis works in AI systems, and why post-incident improvements matter more than a one-time fix.

beginner60 min100 XP

Topic Rewind Recap

Rewind the monitoring and incident response block and apply the core ideas in one defender workflow: review telemetry, write detections, triage the incident, contain the risk, and verify recovery.

beginner45 min125 XP

5

Validation and Secure Delivery

Learn how defenders validate AI features before release, pressure-test controls, and turn security review into a repeatable release discipline.

5 live rooms

Threat Modeling AI Features

Learn how defenders threat model AI features before release by identifying assets, trust boundaries, likely abuse paths, and the controls that should exist before launch.

beginner60 min100 XP

Adversarial Testing and Security Evaluation

Learn how defenders pressure-test AI features before release, what adversarial testing tries to prove, and how security evaluation differs from a generic demo or happy-path QA check.

beginner60 min100 XP

Model, Data, and Supply Chain Risk

Learn how blue teams review third-party models, datasets, embeddings, and dependencies so a feature does not inherit hidden trust or integrity problems from the AI supply chain.

beginner65 min100 XP

Resource Abuse, Cost Controls, and Safe Rollout

Learn how defenders limit resource abuse, apply safe rollout patterns, and reduce blast radius when a new AI feature or model change reaches production.

beginner60 min100 XP

Topic Rewind Recap

Rewind the validation and secure delivery block and apply the core ideas in one defender workflow: review the threat model, evaluate risky behavior, assess the supply chain, and decide whether the feature is safe enough to ship.

beginner45 min125 XP