LLM and Agentic AI Security Engineering
Engineer prompt-injection defences, RAG authorisation, tool safety, agent identity, memory, budgets and abuse tests.
Curriculum
Learn at your own paceTrust boundaries and RAG
0/2 completedA five-question quiz follows this module.
Module quizzes
Use knowledge in a decision
Practice paths for this course
4 missions to explore
Poisoned context: the assistant’s boundary
A document retrieved by a support assistant requests an unexpected tool action. Separate information, instructions and authority.
Revoked access: the index remembers
Source access was revoked, yet AI retrieval returns an old chunk. Trace authority across the source, index and cache.
Delegated agent: the authority chain
An agent calls another tool and loses the user’s scope during delegation. Investigate identity, tool outputs and recovery boundaries.
Queued authority: when is the decision made?
An export request is accepted, then access is revoked before execution. Inspect session, queue and download boundaries.
Final exam
Meet the required proficiency standard to earn your course certificate.
Academy finals require 80%; official exam requirements are assessed separately.