The LLM threat landscape
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
name the main LLM application risks and where they sit in the system.
Planned topics
- The OWASP Top 10 for LLM Applications lists risks such as prompt injection, sensitive information disclosure, improper output handling, excessive agency, system prompt leakage, vector and embedding weaknesses, and unbounded consumption. Verify the current edition and numbering before recording.
- Root cause: the model cannot reliably tell instructions from data.
- Treat the model as an untrusted component inside your system.
- Add the assistant to the STRIDE model from MID-052: user input, retrieved content, tools, output, and cost are all flows.
Draft outline details
Demonstration
DFD of the assistant: user → API → retrieval (vector store) → model → tools → response, with threats marked on each flow.
Check your understanding
Why is the model an untrusted component?
Show answer
It follows text it sees, including attacker text, and cannot guarantee obeying rules.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.